By YEET MAGAZINE | Updated 0200 GMT (1000 HKT) June 6, 2021

Ex Machina - a programmer tests his artificial consciousness algorithms in a robot
"Machines will take the same path as the evolution of human beings, until they become conscious and be able to develop on their own without man being able to understand them or to keep control of them".

This is how Ray Kurzweil describes the next step in the development of artificial intelligence. The character is controversial, he is known for his transhumanist positions in favor of immortality and the convergence between man and machine, but he is also director of engineering at Google.

Another specialist in artificial intelligence and cognitive sciences at MIT, Marvin Minsky also declared that “ nothing could prevent machines from reaching and exceeding human intelligence unless we prevent them” . Added to recent positions taken by Bill Gates , Stephen Hawking and Elon Musk to warn about the advancement of artificial intelligence, this is starting to get serious.

Artificial intelligence is exploding

Conscious machines, it seems unlikely and yet we are at a turning point. Artificial intelligence is taking off. Not a single day goes without a new advance in artificial intelligence being announced. In 2012, Google was teaching its artificial intelligence to recognize a cat by watching videos on Youtube, in 2016 Alpha Go beat the best Go player in the world when we thought it would not happen for at least several decades, Watson diagnosed rare leukemia in a patient in Japan within minutes while doctors were unable to make a diagnosis, in 2017 Libratus was spraying poker players, and it's not about to end. The race for artificial intelligence between the giants of the web is on.

The problem of artificial intelligence is the problem of intelligence itself

This race for artificial intelligence is not new, it dates from the beginnings of computing. And from the start, several approaches have been proposed to create artificial intelligence. First, to imitate the functioning of human intelligence. Problem: we did not know and we still do not know how it works (even if the discoveries are dazzling and they raise new questions).

Then to replicate the human brain itself with transistors or algorithms. There are attempts, but there are still too many unknowns about its organization and its functioning, and even if we have a perfect knowledge of it, it would be too complex and certainly inefficient to use this approach.

And finally allow the machine to perform the same functions as human intelligence (recognize its environment, manipulate language, memorize, contextualize, reason, and many others…) without necessarily knowing how human intelligence achieves it.

The analogy often used in artificial intelligence lessons is that of artificial flight. Artificial intelligence would be to human intelligence what the flight of an airplane (artificial flight) would be to the flight of a bird (natural flight): the two work differently and have two different substrates (biological vs metallic) but achieve the same result, they fly.

Even if it is this last approach which today produces many results, in reality all these approaches combined with research in cognitive science, in philosophy, in psychology,

A reversal of approach that initiates this revolution

This revolution in artificial intelligence, we owe it to a reversal of approach. In traditional computing, software was programmed according to rules to perform tasks under conditions and scenarios well defined in advance. The machine was just doing what you asked it to do.

It is this approach that we wanted to use to create artificial intelligence. Problem: we did not know and still do not know these rules of “functioning” of intelligence.

So we reversed the approach, instead of trying to program machines according to rules to imitate human intelligence, we asked the machine to find the rules itself from examples.

In fact just like babies when they learn to speak: they are bombarded all day with people talking around them, and by dint of being exposed to language, they conceptualize the rules of grammar which then allow them to speak without even knowing they exist. It is thanks to this change of approach that today we can train artificial intelligences to perform very different tasks.

Deep Learning and Reinforcement Learning: two approaches with unlimited potential

From an algorithmic point of view, we use Deep Learning to train artificial intelligences to identify characteristics, high-level concepts in data: it is thanks to Deep Learning that web giants identify what is on photos, it is enough to show a lot of examples to the algorithm and to supervise it by indicating to it what is on the photos so that it is then able to do it on its own on new examples.

Another type of algorithm that is very promising at the moment is Reinforcement Learning. In this approach, we give the algorithm an environment with constraints, capacities for actions and a goal to achieve, and we will let it train through trial and error and exploration to achieve its goal on its Super Mario on his own, and besides he quickly discovers how to play in a totally optimal way.

And when you combine Deep Learning and Reinforcement Learning, you get algorithms capable of both recognizing their environment and achieving goals. It was thanks to the combination of the two that Alpha Go beat the best Go player in the world last year.

But in the future the potential is much more promising since it has now been understood that we can train artificial intelligences to perform any type of task by giving them an environment and an objective to be achieved by training. It's like looking at each goal to be achieved like a video game with a field of view.

Last year the organization OpenAI launched the platform Universe so that everyone can teach artificial intelligences to perform new tasks in all fields (both playing video games and filling out a form on the internet).Tomorrow no one will program machines with code, we can all train artificial intelligences to perform tasks.

Towards a unique artificial intelligence, as a service

One of the many challenges of the next few years is the unification of the different artificial intelligences. Today all the progress in artificial intelligence takes place on very specific use cases (winning at poker, driving Google Cars, diagnosing cancer or replacing managers ), but in the future we will have to create an artificial intelligence able to make the link between all these use cases to allow everything to be done. This is also the objective displayed by Demis Hassabis the creator of DeepMind “ Solve intelligence and use it to do everything”.

This unification is not going to be simple, it will be necessary that the artificial intelligences speak and understand each other, have a global view on our objectives, a working memory, an ability to contextualize, to cut their actions into smaller tasks ... This intelligence unique artificial we will all be permanently connected to it, it will be in a Cloud and it will be our digital butler, it will know us perfectly and will be able to do everything for us, as well as being our doctor, our banker, our lawyer, our advisor, our teacher and our friend who will take care of everything for us and represent us in our interactions: this is the ambition of the web giants with Google Assistant, Facebook M, Alexa from Amazon, Cortana from Microsoft.

Machines will become conscious ... artificially

But as Ray Kurzweil or Marvin Minsky said, this development of artificial intelligence will not stop with algorithms capable of doing what is asked of them. One of the big questions of cognitive science is that of consciousness, and there too researchers are starting to develop algorithms for artificial consciousness .

Again, the same problem as with artificial intelligence: it is difficult to define what consciousness is, researchers do not agree and the findings question everything that we believe (even if many worksoffer very interesting perspectives). So there again we do as with artificial intelligence, we agree on an objective of consciousness and we try to model it in an algorithm. This is what Selmer Bringsjord did, researcher at the Polytechnic Institute of New York.

He defined consciousness as the ability of an organism to observe itself functioning as an entity distinct from others. He was inspired by babies, who around the age of one and a half are able to understand that they are responsible for the words that come out of their mouths when they speak, or recognize themselves moving in front of a mirror. He programmed a robot telling it that he had given it a pill that could either cut it off or do nothing (the placebo).

He then asks the robot what pill he received, and the robot begins to respond to the researcher he doesn't know. And by responding to the researcher, the robot realizes that it is expressing itself and that therefore it has received the placebo: it apologizes and tells the researcher that it can confirm that it has received the placebo. placebo. Two loops of algorithms that work in parallel: one that allows the robot to understand its environment, and one above that allows it to observe itself working and adjusting.

Obviously it is easy to say that it is not human consciousness, that it is only code, and that it is much too reductive for human consciousness… It's true, but it's a start , these are the first algorithms of artificial consciousness. We are obviously far from the film's artificial consciousness tests. and that it is much too reductive for human consciousness… It is true, but it is a beginning, these are the first algorithms of artificial consciousness. We are obviously far from the film's artificial consciousness tests. and that it is much too reductive for human consciousness… It is true, but it is a beginning, these are the first algorithms of artificial consciousness. We are obviously far from the film's artificial consciousness tests.Ex-Machina . But the trend is on and research continues.

One of the first very simple artificial consciousness models made at the New York Polytechnic

Be inspired by the living

The trend in the development of these intelligence and artificial consciousness algorithms is to draw inspiration from the living . In the same way that DNA can replicate, mutate, merge and be transmitted to evolve, the algorithms will be inspired by these principles in order to be able to replicate, mutate and be selected by their environment for their ability to reach their goals. Goals. Obviously, by giving them these properties, these algorithms will become self-complex by evolving until they become potentially incomprehensible for their creators, and this is the risk of loss of control that researchers like Stephen Hawking or Marvin Minsky talk about.. Needless to say about sci-fi or end-of-the-world scenarios, we have no idea what these self-learning, self-organizing algorithms will look like. Moreover, we are already having trouble understanding the behavior of Deep Learning algorithms .

It's never going to happen !

“An artificial intelligence capable of doing everything? Impossible, never will we achieve such a complex and flexible intelligence in a machine! ”,“ We ​​will always need man for… this or that… ”,“ The machine can never be creative! ”,“ The artificial consciousness is science fiction ”… This is what we are hearing right now and will continue to hear in the coming years about artificial intelligence. Even if the prowess of artificial intelligence explodes, there are still journalists, philosophers, consultants and even researchers to tell us no, general artificial intelligence, the singularity point where machines might be able to to develop on their own, we will never or will not succeed for a very, very long time. But there are two arguments for thinking they are wrong.

The first is the exponential speed of the development of artificial intelligence. In the same way that genome sequencing was made possible and democratized in a few years, artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years.

Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled. every 18 months which began to The first is the exponential speed of the development of artificial intelligence.

In the same way that genome sequencing was made possible and democratized in a few years, artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years.

Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled. every 18 months which began to The first is the exponential speed of the development of artificial intelligence.

In the same way that genome sequencing was made possible and democratized in a few years, artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years.

Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled.

Every 18 months which began to In the same way that genome sequencing was made possible and democratized in a few years, artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years.

Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled.

Every 18 months which began to In the same way that genome sequencing was made possible and democratized in a few years, artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years. Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled.

Every 18 months which began to artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years . Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled.

Every 18 months which began to artificial intelligence and all other scientific and technological advances are currently following this pattern of exponential acceleration which makes them very close to us: what we thought could be done in 20 years will probably be in 5 or 2 years .

Our brains imagine the future in linear fashion as technology evolves at an exponential rate, making this phenomenon hard to imagine, but it is very real (even with Moore's Law that the computing power of machines doubled. every 18 months which began toslow down and start again …).

The second argument is philosophical, that is to say that any phenomenon that we define can be modeled in an algorithm and that it becomes this phenomenon… in an artificial version.

The difficulty lies more in the definition of these phenomena and in the fact that they evolve at the speed of discoveries which is also accelerating. Humans also tend to think that each time the machine manages to do a task that we considered to be intelligence, in the end, it is not intelligence, since the machine has done it!

It's moving fast, but it's neither simple nor magical

The recent results of artificial intelligence are impressive and the prospects are dizzying. But it should be remembered that these results are the fruit of considerable work.

Deep Learning does not learn to recognize what is in photos on its own, it needs to be supervised by a human being to guide it, clean data that also takes a long time to be ready to be. processed, programmers to configure the algorithms and choose them, considerable computing capacities…

The coming years will provide us with solutions and optimizations to all these obstacles. Discoveries about the brain will inspire us to develop increasingly powerful and flexible algorithms, and the behaviors of these algorithms will also enlighten us on the functioning of our own intelligence. There are going to be obstacles (beliefs, interests, lobbies, public reception, questions relating to privacy and data, policies, etc.) and we will have to prepare the framework for the organization now. artificial intelligence from an ethical and legal point of view to a global level.


A collection of 15 posts

10 photos of Facebook CEO, Mark Zuckerberg’s California Home


10 photos of Facebook CEO, Mark Zuckerberg’s California Home

When you think of Mark Zuckerberg (and his wealth and global status), how do you imagine his home to be? An OTT space, robots for house help, possibly a helipad at home? Well...that visual isn't entirely true.



News -US - Brain implants allow you to “ taper ” with your mind


News -US - Brain implants allow you to “ taper ” with your mind

Scientists at Stanford University in the United States have developed a «brain-computer interface» (BCI) which converts mental writing into words on a screen.



Technology /#Drone #IA #Artificial Intelligence : How This Tesla Was Hacked By A Drone In Minutes! Demonstration.


Technology /#Drone #IA #Artificial Intelligence : How This Tesla Was Hacked By A Drone In Minutes! Demonstration.

Hacking cars from the sky might sound like something out of a sci-fi movie like Fast & Furious , but researchers have shown how close this technique is to reality. They



Artificial Intelligence: Microsoft Ready To Buy The Artificial Intelligence And Voice Technology Company Nuance Communications Inc. For $ 16 Billion


Artificial Intelligence: Microsoft Ready To Buy The Artificial Intelligence And Voice Technology Company Nuance Communications Inc. For $ 16 Billion

Microsoft is said to be on the verge of signing its largest acquisition since that of Linkedln in 2016. The transaction could be announced this week.



Smartphone Business During The Pandemic: How Table turns For Micromax In India, As Chinese Manufacturers  Get  Axed By The Popular Indian Cell Phone Brand.


By Salil Trivedi



Elizabeth Holmes and Theranos, Silicon Valley's biggest scam.


Elizabeth Holmes and Theranos, Silicon Valley's biggest scam.

She was the youngest self-made billionaire in history. She was to become the next Steve Jobs - she had borrowed his turtlenecks from him. At just 30 years old, Elizabeth



Tech News: Coinbase CEO Brian Armstrong received $1.78 million to cover “personal security” expenses in FY2020.


Tech News: Coinbase CEO Brian Armstrong received $1.78 million to cover “personal security” expenses in FY2020.

Coinbase CEO Brian Armstrong received $1.78 million to cover “personal security” expenses in FY2020 as the 38-year-old tech CEO  prepared to take the cryptocurrency exchange operator public. Armstrong, who



Trending: Could A Call With A Stranger Help You Concentrate? These App Makers Think So.


Apps that can make you money? It sounds too good to be true, but moneymaking apps do actually exist. While these apps won't make you rich, they can make an



An artificial intelligence imagines the "real" traits of historical figures


An artificial intelligence imagines the "real" traits of historical figures

An artificial intelligence uses deep fake to recreate  the "real" traits of historical paintings.When artificial intelligence brings great historical figures to life. When artificial intelligence brings great historical figures



Apple's new iPhone and iPad designs are like nothing we've seen before


Apple's new iPhone and iPad designs are like nothing we've seen before

At present, the iPhone's rear camera is much more powerful than the front lens, but an extra screen could make it possible to use the back camera instead.






Snapdragon 855 Plus processor, quad-camera setup for RM2,399 Flagship smartphone also includes 50W super-fast charge (From left) realme Malaysia brand manager Tiffany Teh; realme Malaysia CEO Hunter Wang; Qualcomm



Spotify goes to war with Apple


Spotify goes to war with Apple

On March 13, Spotify filed a complaint with the European Commission that accused Apple of violating antitrust laws. Spotify CEO Daniel Ek summarizes the critique in a blog post, writing:



This Dad Sold His App For $54 Million and Went Around The World!


This Dad Sold His App For $54 Million and Went Around The World!

What would you do if you became a multi-millionaire overnight? How about sell all your belongings and set off on an incredible adventure with your family in tow? That’s



Apple to release three phones: Report


Apple to release three phones: Report

Apple to release three phones: Report In the increasingly competitive market for smartphones, Apple is working on releasing three new iPhones later this year, Bloomberg reported. One, code-named D33, will



Meet Harmony, the sexy sex robot that can snatch your boyfriend from you.


Meet Harmony, the sexy sex robot that can snatch your boyfriend from you.

As experts believe that by 2050 men might be getting married to sex robots, here is Harmony, the frontrunner in the race to become the first commercial sexbot who has




Can an artificial intelligence create another artificial intelligence ?

François Désarménien

It all depends on what you mean by " artificial intelligence ". For example, I wrote several computer programs whose function was to generate other programs: that does not pose any particular problem.

There are a lot of phantasies surrounding the expression AI coming from the fact that the term " intelligence " in English also means espionage / intelligence / gathering of information and that it is this meaning which inspired the use of the term to l 'origin for computing.793  viewsShow 7  positive votes ·

How to create artificial intelligence yourself ?

Fred cavalier

How to create an artificial intelligence yourself?

This is a very good question because, yes you can create an artificial intelligence yourself . If you mean by artificial intelligence , a robot capable of having a conversation, know that it will be very long… and complicated, unless you have an army of researchers by your side. Nevertheless, it is quite possible to design artificial intelligence programs that meet certain rules such as for example to do data analysis or prediction (a concrete example of data analysis and processing:

A AI made by makers turns Nic Cage into James Bond. ).

Today market learning and deep learning, the creation of artificial intelligence programs , are available to everyone on the internet, with courses on AI, code examples, complete projects. Concretely, anyone in their living room can design an AI - with a lot of method and perseverance however).

It all depends on your computer level and your willpower.

If you don't know anything about code, I recommend you to learn python language, ideal for artificial intelligence programs and a language easy to master when starting out.

KDnuggets, a site in English, specializes in data science, machine learning and big data. You will be able to read very good articles on the subject, as well as complete tutorials.

While searching a little on the net, you will come across many AI application projects made by enthusiasts like this page for example: 8 Fun Machine Learning Projects for Beginners .

So, if you want to give it a go, it's possible!

How to create an artificial intelligence that plays chess?

Thomas seze

Studied Cognitive Sciences and Artificial Intelligence at the University of Bordeaux 1 ( Diploma obtained in 2016 )Original answer: How to create an artificial intelligence that plays chess?

There are several possible methods, I will outline the two that I think are the most interesting:

  1. Use of the Q learning algorithm , the most complex solution but which will give rise to the most “intelligent” AI.
  2. Use of the min-max algorithm , the simplest solution but requires a lot of calculations (because it must calculate all the possibilities on X turns).

In both cases , 2 elements are required:

  1. The state of the environment (here, chessboard and the different pieces with their positions)
  2. The list of possible moves (here, each piece does not have the same characteristics)

Let's start with simplicity:

The min-max algorithm consists of maximizing gains and reducing losses. To do this, it calculates all the possible moves for each piece present in the current game.

The algorithm itself is very simple , 5 or 6 easily understandable lines of code.

Just assign points to each piece:

Relative value of chess pieces - Wikipedia

Example: pawn is worth 1, bishop is worth 3, etc.

The algorithm calculates all the moves and assigns a value, to eat a bishop = -3, eat a queen = +9.

After analyzing all the possible moves, the algorithm will bring out the best move which leads to the best scenario.

It remains to configure the depth (the number of moves over time), because if in the current state action X brings 3 points, but that the next time the best move will also be 3 (6 in total), the An action is less interesting than one that earns 1 point this turn but 9 the next (10 in total).

Thus the depth will give the “intelligence” of the algorithm, but also exponentially increase the computation needs.

This solution is very quick to implement.

Note that this solution is not machine learning, but remains an AI.

Now Q learning:

This is the algorithm of choice used in the biggest projects (the ones that beat the best chess / go / etc players in the world).

However, if the concept remains relatively the same as the min-max, maximizing the reward (reward = sum of positive points - sum of negative points), understanding the algorithm is not trivial, but remains understandable with a little of effort.

But the most difficult remains the implementation. translating the function into algorithmic form and adapting it to the desired case can prove to be long and complex.

How it works :

At the very beginning, the algorithm does not know anything, it will try random moves. Little by little he will learn basic concepts, eating is good, being eaten is bad. And as the games go on, he will learn more complex scenarios (Machine learning) “You have to develop the riders in such and such a place while freeing up space for the madmen to maximize the number of moves possible”, etc.

Thus, the more the algorithm will play games, the better the AI ​​will perform, just like humans, with the difference that on a good server, the algorithm can simulate more than 100 games per minute, whereas 'we need 30 minutes to make one.

The effort is worth it because it is the solution that will give rise to the most developed intelligence .

To summarize , if the goal is to do a small personal project => min-max

if the goal is to make the most efficient AI with less computation => Q learning.

If the goal is to learn to code an AI, the first then the second.6.2k  viewsShow 42  positive votesShow shares ·

What do you think of the "resurrection" of loved ones by artificial intelligence ?

Profile picture for Pierre Leclercq

Pierre Leclercq

Can artificial intelligence develop feelings for its creator?

Profile picture for A-Alexandra Boca

A-Alexandra Boca

We must first manage to contain our feelings in a test tube in order to study them, model them and learn them by machine.

The day when we have “dissected” happiness or hatred, we will be able to supply them to the machines and see if they are capable of reproducing them, but again that would mean that we love or hate all in the same way.177  viewsShow 3  positive votes

Is it possible to one day be governed by an artificial intelligence ?

Profile picture for Jose-Marcio


PhD in Artificial Intelligence and IT Security , MINES ParisTech ( Diploma obtained in 2011 )

Maybe yes, maybe no ...

In "2001, a space odyssey" , the HAL computer took power… or rather, it was humans who put it in this situation or let it happen… So, it would have to be this either humans who delegate, entirely or not, the task of governing, or that, by exploiting a loophole, they can appropriate power.

For those who have not noticed the meaning of the HAL name assigned to this computer , replace each letter with the next one… and you will have… :-)

But more concretely, the points that I see ...

From a philosophical point of view, it seems hardly acceptable to me to be governed by a computer. Unless we come up with scenarios like "1984" or "Brave New World" . As it stands, I find it hard to see how an AI could have a dialogue capacity similar to that of a human (sometimes nil).

From a technical / scientific point of view, an " artificial intelligence " is always a couple: " processing intelligence " + "knowledge ".

The intelligence process is a set of algorithms capable of studying the relationship between a new situation presented and similar situations and consequences, from the "knowledge" (or basic).

And that is the problem. The "knowledge", are the initial data: the history books, the press, ... possibly updated regularly.

So, for a given problem, the solution proposed by the AI ​​will probably be the best it can find in its "knowledge". He is not sure that he can innovate, that is to say, find a solution that is not in his "knowledge" .

On the constitution of its "knowledge" (in fact, in AI, this is called learning ), it is necessary to make sure that there are enough situations, historical, with good decisions than with failures, so to make the best decisions and not the least bad ones.

Again on the constitution of “knowledge” , the reliability of sources is very important. When we see the amount of f ake news delivered by a biased press or, worse, parts of history being rewritten, this “noise” must be negligible. If you take a kid with an IQ of 500 and raise him in a thug environment, don't wait until he becomes a wise man.

Another situation that comes to my mind are "rare events", that is to say situations which have only appeared very rarely over time, and whose results are not statistically significant.

But… after all… with the leaders who are everywhere… one can wonder if, despite all these limitations, an artificial intelligence would perhaps be able to do better than the humans who govern us… (some smileys…).

The question is super interesting! I am certainly not exhaustive.

Thanks Gérard.

Can we create artificial intelligence with PHP?

Khalid Moustapha Askia

Master in Artificial Intelligence and Mathematics , University of Quebec in Montreal (UQÀM)

PHP being an object oriented programming language (despite these drawbacks), it is certainly possible to do some artificial intelligence algorithms with it.

Unless you are in danger of death and PHP is the last solution to survive as much avoid.

How can artificial intelligence evolve?

Louis Philippe Doyon-Lessard

software developer

By using an evolutionary algorithm, it is possible for an artificial intelligence to improve its performance.

As with natural evolution, genes are assigned to individuals corresponding to characteristics that must be optimized . These individuals are made to evolve in an environment iteratively. At each iteration, we keep the best performing individuals. It is the survival of the best suited individual. Also, the genes of the best individuals are used to create similar offspring, but with some mutations. The more time passes, the better the individuals of the simulated population are.

Karim Said

How do you create artificial intelligence ?What would artificial intelligence look like ?

Profile picture for Alain Coetmeur

Alain Coetmeur

I imagine that by the term ' Artificial Intelligence ' you mean what is called ' Artificial Intelligence Generalist "(AGI). This is not won because current AIs are very specialized components that can be misunderstood as soon as they are used where they have not been trained.

They are seriously mistaken for example in the prediction of orders in supermarkets, during surprise lockdowns. And we had to redesign them to adapt to faster changes, and feel the trends of the moment and more habits. A weak AI as we know it said, is content to learn to behave as one expects of her, in a specific domain (face, traffic prediction, musical preference, consumption), according to the data provided to her, and l

What is the best computer language to create artificial intelligence ?

Profile picture for Gregory Popovitch

Gregory Popovich


No one knows how to create artificial intelligence , so there is no definitive answer. But still, it is likely that a real intelligence must be able to modify its own program, so perhaps a language like Lisp would be the most plausible.  

How to create artificial intelligence like Alexa?

Alex Coninx

Lecturer in computer science at Pierre and Marie Curie University ( 2016 - today )I would put aside the question - as usual uninteresting and moving - whether Alexa is "really AI" and focus on the heart of the matter. There are many free projects that allow you to build a personal assistant who responds to the voice. These projects are generally deployable on a single board computer (Raspberry Pi type), and operate more transparent and more respectful of privacy than those of large multinational IT companies. However, some use, generally optionally, certain components that they provide (often in particular Google voice recognition), which then exposes the corresponding data.

A quick tour of the projects I found.

What difference is there between intelligence artificial specialized and intelligence artificial general?

Profile picture for Siméon Ntakarutimana

Simeon Ntakarutimana

Former paramedical student ( Diploma obtained in 2013 )

Specialized artificial intelligence is understood to mean techniques and applications intended to accomplish single or restricted tasks, while “ general artificial intelligence ” or “superintelligence” refers to systems capable of satisfactorily performing any intellectual task.

Do you find it relevant to qualify the result of artificial intelligence as art ? Why ?

Profile picture for Natacha Natinat

Natacha Natinat

Lausanne, VD, SwitzerlandAnd why not ??? Ironically, I go back to 1945 to find an angle of answer to this question on AI ... Do you find it relevant to qualify a page of a diary as art or the juxtaposition of words written on a piece of paper ( not intended in the eyes of others) by a resident of psychiatric asylum without notion or artistic intention? Those words in fly's feet linked above without ever any space, woven frantically by the so-called Justine Python (1932) - poor lady who had no light on all floors - were quite impressive to discover as they covered an entire wall, during an exhibition at the Collection d'Art Brut museum in Lausanne. In 1945, the French painter Jean Dubuffet, still almost unknown but above all excess

Does France really have the best researchers in artificial intelligence ?

Profile picture for Daniel Ribeiro

Daniel Ribeiro


Far from there.

We have some of the best AI engineers in the world, but they all go overseas. A good engineer earns between 40 and 60,000 € per year in France but you just have to cross the Atlantic and go to Canada for you to be offered at least three times, you will find a job in a few days in Canada.

We have good schools and institutes but the budget allocated by France for AI is only a few million while China devotes billions to it. In 2017, $ 15.2 billion was invested globally in start-ups specializing in the sector. And nearly half of that went directly to China, compared to 38% to the United States. In 2018 the gap widened further. China is therefore the world leader in the field of AI.

In the United States, the typical specialist in artificial intelligence is generally a doctor who has just graduated from university, or if he is less qualified, with barely a few years of experience. He is paid between $ 300,000 and $ 500,000 a year or more, in salary and shares. For the references of the sector, the package can be negotiated several million dollars, for contracts of four to five years, which they can then renew like a professional sportsman. A relevant analogy, since some even say, half-jokingly, that ceilings should be established in the image of what exists for players in the National Football League .

Even at the European level, the budget is ridiculous, if there is no immediate jump, we will become the slaves of big countries like the USA and China and when India wakes up, we will only have to cry.2.2k  viewsShow 6  positive votes

What is the relationship between mathematics and artificial intelligence ?

Profile picture for Philippe Louis Delacan

Philippe Louis Delacan

Master in Machine Learning and Imaging , École des Ponts ParisTech (ENPC) ( Diploma obtained in 2018 )Artificial intelligence contains several branches related more or less to mathematics, in particular but not exclusively: * Machine learning or machine learning, is largely based on mathematics (in particular statistical learning, learning by reinforcement and evolutionary models). * Logical reasoning is based on logic and algebra * Signal processing (Images, Sounds, etc.) relies almost entirely on mathematics * Computational neuroscience models use mathematics * Graph theory has some links with mathematics but is above all a branch of algorithmics.

What is the average cost of an artificial intelligence project ?

Profile picture for Thomas Seze

Thomas seze

Studied Cognitive Sciences and Artificial Intelligence at the University of Bordeaux 1 ( Diploma obtained in 2016 )

To resume the linked article:

“It goes without saying that the costs vary enormously from one project to another. “

I can nevertheless provide a few examples to give an order of magnitude:

  • Project: Automate document processing (item style, invoices, excel file, etc.)

Salary of a junior data scientist 1900 euros NET per month (depending on region), which costs the employer approximately 3400 per month. It will take him (depending on his intelligence , his skills, the complexity of the environment, etc.) more or less 8 months to complete the project. 3400 * 8 = 27,200 .

27,200 euros to set up the system, this does not include maintenance and long-term bug fixing etc.

  • Project: Automatic retrieval of potential buyer profile for E-commerce / Suggested item for sale according to a customer profile

Allow about 4 months if the web structure has been well thought out initially.

4 * 3400 = 13,600 .

  • Project: Automatic driving learning for robot (Arduino type)

It takes 2 Data Scientists for this task, to diversify points of view and speed up the process. And about 15 months to finalize the project.

2 * 3400 * 12 = 102,000

It should be kept in mind that these examples are small projects such as startups or medium-sized companies, I took junior data scientists as a reference, I did not include in these calculations any training or equipment purchase costs, etc. And we can still see that the costs soar very quickly, but the projects are nevertheless enormously profitable in the long term.

In addition, as said before, the complexity of the environment can greatly influence the duration of projects (and therefore costs). If it is difficult to recover the data, that they are insufficient in number, that the process to test the new algorithms is long, that each modification it is necessary to fill out papers and make a report, we can double the time design / cost .

Now if we imagine a large-scale project like at Airbus, with an entire team of data scientists, equipment blows etc, it can quickly rise to a few million .

(Paragraph by pure anticipation)

The duration estimates take into account the real work processes in the company , with the design / development / extensive test phase / production / bug correction, integration into an existing project, understanding of the existing one, etc ...

Obviously, a data scientist can design functional algorithms in 2 days if all the data has already been collected / sorted / cleaned / analyzed, that there are no language / performance / scalability / code quality constraints, etc. … But in the real world it doesn't work like online course tutorials.

Is Siri an artificial intelligence ?

Profile picture for Richard Trigaux (Yichard Muni)

Richard Trigaux (Yichard Muni)

Location: Rodez, France

Siri, I don't know.

But Cortana, the Microsoft equivalent, no.

Cortana is just a database of ready-made responses, reacting to keywords. If you ask her a question she doesn't know, she opens the Bing search engine. Siri can't be any better.

Here is Valaba real artificial intelligence , created in 1991 by one of my work colleagues at the time. This stuff is able to "sign off" laboratory tests for a doctor, with more certainty and fewer errors. And for that he does not need a sensual female voice or science-fictional blah. If I remember correctly, he was using Prolog, a high level logical inference engine.

If I want talking artificial intelligence , this is its voice:162  viewsShow 2  positive votes ·

How can we maintain control of an intelligent and complex system ( artificial intelligence )?

Profile picture for Dembele Jean-Marc

Dembele Jean-Marc

DeveloperI think we overestimate AI a bit today. So yes, an AI can go nuts, and cause serious damage. But these AIs, which can potentially get out of hand, not because they've gone wild (literally) but because it's a side effect of their learning, or a bug, are theoretically still under control, at a physical minimum. If a “problem” were to arise, it would be human error, or malicious intent on the part of its creator. We do not create an AI that collects personal data from users of a social network, while teaching it how to use this data to improve and / or harm those same people, and at the same time, educate it on how to protect yourself from an external (human) shutdown, in the code as well as in the physical(more)

Artificial intelligence in the medical world: good or bad idea?

Profile picture for Tristan Roth

Tristan roth

I beg my friends to start their plans

Artificial intelligence presents many opportunities for the medical world:

  • Prediction of a disease and / or its course
  • Personalized treatment recommendation
  • Diagnostic and therapeutic decision support
  • Companion robots for the elderly or the frail
  • Computer assisted surgery
  • Prevention

However, there is at least one bottleneck: the quality of the data sample. To be effective, the digital approach requires clean and well annotated data , which is rare. The largest French database, SNIIRAM, was not designed for medical analysis, but economic analysis.

Another challenge is to be able to exploit textual information about patients . There is much hope for this in unsupervised learning algorithms, but for now, they still require a lot of research before they can be used in any meaningful way.

Other questions also arise:

  • how can AI deliver the right information at the right time and to the right person , but also deliver it in a form that can easily be understood.
  • how can AI avoid reproducing the biases and biases that were present in the learning data it was trained with.

Alain Hollemaert

Should human rights be granted to Artificial Intelligence (AI)?

Profile picture for Robert Lafite

Robert Lafite

Electronics development technicianIf an artificial being is capable of feeling positive or negative emotions, there should be a law for his interests to be taken into account. This kind of law exists for animals, although they are very poorly enforced. However, it would be problematic to differentiate a real feeling from a pure and simple simulation. For example, it would be very easy to program basic software that would claim to be fearful of death, and beg us not to shut it down.

What is Deep Blue ( Artificial Intelligence )?

Profile picture for Jean-Philippe de Lespinay

Jean-Philippe de Lespinay

Awards.AI 2017 Prize for Maieutics and Reasoning AIDeep Blue is not artificial intelligence at all ! These are algorithmic programs written with computer languages ​​by computer scientists. We are far from it. Intelligence can play any game. Deep Blue is unable to play Go or poker. Everything would have to be reprogrammed from scratch. These AI pseudo-programs cheat to win: they calculate. While the man, he reasons to play. This is intelligence. A chess player (go, poker, jeopardy, etc.) absolutely cannot compete with a computer that has memorized all the starts and ends of games played for centuries and calculates millions of moves in seconds to retain the most winners. And yet humans have managed to beat this AI (Gary Gasp

What are the best artificial intelligences capable of?

Profile picture for Basil Starynkevich

Basil Starynkevich

You have to give your definition of artificial intelligence .

For example, was Jacques Pitrat an actor in artificial intelligence in France?

Can artificial intelligence discover the meaning of life?

Christian Cappelle

Formerly Teacher and nature guide"This is one of the most asked questions of the AIsphere. Today the majority of scientists think it necessary to integrate some form of consciousness into AI so that it is able to equal human intelligence. . And rightly so. If this question cannot be formally answered, because it is difficult to define what consciousness is (if it is even possible), I will try to offer you a draft analysis which does not 'is in no way formal proof but a personal approach to the subject. " Extract from the article of: Eytan Messika - Expert Contributor Consultant and teacher in artificial intelligence . Founder of the blog specialized in AI inside ai minds. As a reminder, the points of view developed here are strictly personal. Artificial Intelligence Will T-She A j

Which country is the leader in artificial intelligence ?

Aunt Florent

Web and mobile developer ( 2015 - today )

Country Japan is a leading country in artificial intelligence

What is artificial intelligence ?

Profile picture for Sebastien Dourlens

Sebastien Dourlens·

partner at Witty Knowledge ( 2013 - present )I agree with Mr. or Mrs. Anonymous who describes the term IA as used today and not in its entirety but in substance everything it presents is correct. An intelligence artificial is intelligence built by man, and therefore unnatural. Intelligence is not really defined, it is the sum of the capacities or the successes to carry out tasks (mental, physical or others) which defines intelligence, as an observable of an underlying thought in an object " alive "(an artificial life or real). For me and for the researchers of the 1900s, AI corresponds to all the sciences constituting a foundation of "human" intelligence which today is called cognitive sciences for the representation of the environment, of thought, of human beings. emotions

To what extent could the state be replaced by artificial intelligence ?

Profile picture for Mohammed Grini

Mohammed grini

Former Auditor at Algeria Post, retired ( 1976 - 2010 )China is already getting down to governing its people through the data it collects through the web using AI In other words, it does not need this democracy that the West boasts of. See detailed article here:

Is Excel an artificial intelligence ?

Profile picture for Constance de Quatrebarbes

Constance of Quatrebarbes

Ingenious at Data Science

No, and besides, no software or technology could claim as such or in a purely marketing sense of the proposition "to be an artificial intelligence ".

(For a definition of artificial intelligence I invite you to read this:

Intelligence artificial)

Let me explain:

Artificial intelligence proceeds from algorithms, i.e. sequences of instructions (commands) expressed in an unambiguous language that the machine understands and executes, in order to automate a specific task (hence the name of programming) some of which can simulate human behavior (reasoning, learning to talk).

Excel is a tool, it has in hand a programming language (Visual Basic) which allows you to automate certain actions such as filling in boxes and doing calculations but 1. it will not be able to do it on its own 2. It will not not simulate human behavior.

an artificial intelligence easily win the lotto?

Profile picture for Lepine Kong

Lepine kong·April 19

Ex Entrepreneur / Ex ProTrader | Dev Engineer / PO / PM ( 2021 - present )

Why ? She acquired the gifts of Nostradamus too

Is it a language error to speak of artificial "intelligence" ?

Profile picture for Silver Thomas Ayissi

Silver Thomas Ayissi

Web developer

When we talk about " artificial intelligence " we must always keep in mind that it is a word coming from the jargon specific to a profession .

A "hat" in the clothing industry and a "hat" in journalism (defined here: press glossary, press dictionary, typography, journalism) mean two completely different things.

Likewise, it should be remembered that intelligence in the "IT" sense of the term is not the intelligence that we appreciate (or hope!) In a child as a parent or teacher. .

The Larousse dictionary defines artificial intelligence as

The set of theories and techniques implemented in order to produce machines capable of simulating human intelligence.

When you use voice recognition, facial recognition or more simply input assistance in software that allows you to enter text, you are indeed using artificial intelligence in the sense of the IT sector .

When a car parks automatically, it is indeed artificial intelligence in the sense of the jargon of the IT sector.

The Bmw that parks by itself .... - Dailymotion video

Artificial intelligence is very present in the aeronautical field as reported by this Belgian media An airplane lands in total autonomy for the first time (video) - Geeko when he explains:

Researchers at the Technical University of Munich have developed a system that allows small planes to land without this ground aid. This innovation uses light signals and infrared cameras for orientation. The on-board computer is able to recognize the track thanks to the system which guides the machine.

In short, talking about artificial intelligence is not a language error . Artificial intelligence is a reality that is already improving our lives on a daily basis.

The term " artificial intelligence " belongs to the jargon of the IT sector and should be understood taking into account that it is a term specific to a specific discipline . It is also this understanding taking into account the belonging of words to professional jargon that you have when you keep in mind that an "injection" in the mathematical sense is (radically) different from an "injection" in the mathematical sense. medical.

Can artificial intelligences one day feel emotions?

Profile picture for Mikael Koutero

Mikael koutero

Data Scientist

She will not be able to feel emotions in the biological sense of the term, but it is likely that we can develop a way of mimicking a feeling and its consequences.

I know that there are reflections / tests that are carried out to find out if it is interesting for an AI to feel fear. Part of the challenge is to know if, thanks to fear, we can succeed in instilling the notion of prudence and respect for others in an artificial system. The logic being that a large part of our societal behavior is driven by fear, starting with the fear of ridicule. But this mainly concerns only the so-called strong AI. Weak AI, capable of limited tasks, hardly beyond a defined perimeter, is not affected.

The question is not necessarily will they be able? But will it be of use? If so, I think humanity found solutions in this direction.

What does an artificial intelligence researcher do ?

Profile picture for A-Alexandra Boca

A-Alexandra Boca

Artificial intelligence consultant at IBM (company) ( 2018 - present )

Artificial intelligence has several sub-domains: NLP, computer vision, audio, expert systems. The research can be in a fairly precise subject in one of these sub-domains such as for example speech to text optimization or text classification algorithms.121  viewsShow 1  positive vote ·

What was the first artificial intelligence ?

Profile picture for Abdelaziz Elkfita

Abdelaziz Elkfita

Formerly Fleet manager, mecanical ingenieer, author, investorI think fire ancient civilizations changed fire to move their lives agriculture and seeds

If an artificial intelligence decided to destroy all human life, how would it go about it?

Gerard Bigot Paris, FranceIf an artificial intelligence decided to destroy all human life, how would it go about it? This is the premise of Terminator, the movie. In this universe, AI uses atomic bombardment to eradicate humans. There is just one problem with this: the nuclear weapon produces an electromagnetic discharge which destroys the computers, which are the supports of the aforementioned IA. I don't think these AIs would commit such suicide. Rather, I propose to make the environment unlivable for humans, by destroying the biological ecosystem with which these famous AI have nothing to do. It is certainly longer, but it is safer.

What is the difference between software engineering and artificial intelligence ?

Formerly Technical and General Director, CEN TC, AFNOR, LNE

good night Michel

They are two totally different things and not comparable:

Software engineering is comparable to civil engineering, engineering in the military.

Artificial intelligence is a program that seeks to imitate the intelligence of a human being, but our technology is still very far from developing a program with faculties close to those of a human, and fortunately, because they would be able to judge us and act accordingly 🤫


Would real artificial intelligence kill humans?


PhD in Artificial Intelligence and IT Security , MINES ParisTech ( Diploma obtained in 2011 )

No ! It would be illegal !!!

Do you know Isaac Asimov's three laws of robotics? It dates back to 1939, but we talk very little about it in recent years, when we talk more and more about artificial intelligence .

  1. a robot cannot harm a human being, nor, by remaining passive, allow a human being to be exposed to danger;
  2. a robot must obey orders given to it by a human being, unless such orders conflict with the First Law;
  3. a robot must protect its existence as long as this protection does not conflict with the first or second law.

The Three Laws of Robotics page - Wikipedia worth a visit for those who don't know or have forgotten about it.

Can artificial intelligence be multiple?

Profile picture for Jose-Marcio


Hmmm… what do you call “multiple” ???

I will answer on the way that I understood the question.

Roughly speaking, all artificial intelligence (machine learning) devices are created to deal with one and only one specific problem. Sometimes you have systems that can handle multiple problems, but they are all known in advance and it's more of an integrated system with as many methods as problems.

An AI device capable of dealing with an open, new problem… I think we're still a long way off (just a personal opinion).

Also, I'm talking here about “AI device” (or similar term) rather than “artificial intelligence”.

It's not the same if you get picky.

What does the code of an AI ( Artificial Intelligence ) look like?

Profile picture for Jean-Bernard Francois

Jean-Bernard Francois

operating systems expert

The code looks like a sorting algorithm

Mathis Valdant

Can artificial intelligence be creative?

Profile picture for Michel Verheughe

Michel Verheughe

Internet user since 1994

One day I read that we had tested this by people who "spoke" with something, not knowing if it was a human being or a computer.

Several made a mistake because, this computer, not including / understanding what was said, answered by a sentence of the kind "but what do you mean by that?" And it seems that answering one question from another is very… human.

My answer will therefore be: Yes, artificial intelligence will be creative if it is programmed to be so. But perhaps we should first of all define the meaning of creativity.

What is Artificial Intelligence used for ?

Profile picture for François Paganel

Francois Paganel

Work at IBM (company)

Right now to outsource the simple issues that bother people. In the short term to help people. Longer term to replace people.

What are the main misconceptions about AI ( artificial intelligence )?

Profile picture for Laurent Bourrou

Laurent Bourrou

AI - Semi-amateur cogniticianThe one that AI removes from employment. “Mon” IA is the world leader in predicting failures on expensive industrial equipment. We can save our customers millions of dollars just for ONE failure avoided. And we have rather helped to create jobs than to lose them. Our customers still have their old experts in the field, but they now also have young people in control rooms thousands of kilometers away calling them to ask them why a certain ball bearing is 2 ° hotter than yesterday. And the old experts answer “who cares about 2 °”. And the young people say “Nah. My model predicts it will blow in 3 weeks ”. And where the story is beautiful is that the famous models need the old experts. AI is very human in

Will robots with sophisticated artificial intelligence be able to replace artists in the future?

Profile picture for Flavien Volken

Flavien Volken

Studied at École Polytechnique Fédérale de Lausanne

This is already more or less the case today Vincent, the AI ​​who transforms sketches into works of Van Gogh or Picasso - Tech - Numerama

But these algorithms for now only imitate us. Admittedly, they recreate something new but on axioms which they have learned by imitation of a known domain. It is very difficult to be creative, much more than you can imagine. The brain itself is already struggling to be so. For example if I ask you to imagine a new color that comes out of the visible spectrum, I don't think you can.

Now imagine that AIs manage to do what artists of today manage to do, the new challenge will then be to make art that robots cannot imitate. There was a time when cameras did not exist, the ultimate painting was to make a very faithful reproduction of reality. As soon as cameras appeared, realistic paintings lost their interest.

We have a very big advantage on the machines that for the moment, they are not ready to have access: The awareness of what is pleasant and what is not. To be well trained, an artificial intelligence algorithm would need to have this feedback on whether a stimuli is pleasurable or not. However, on the one hand we do not even understand ourselves why certain things we like more than others (good luck explaining it to a computer) but above all it would be necessary for a computer to have some kind of physical presence and probably even spiritual to really be able to train. So it's still a long way off… unless we train algorithms on the basis of human stimuli, but there again… they would imitate us.

Sylvain Saurel

How many jobs will be lost after the arrival of artificial intelligence ?

Profile picture for Jean-Marc Nahon

Jean-Marc Nahon

Research & Innovation Director at GESTFORM

Germany is one of the countries with the most robots and yet there is little unemployment.

The AI ​​will lose jobs, software robots too (RPA), but they will create them too. The problem is that the qualifications evolve, 70% of the positions to be filled in 5 years will be in trades which do not exist today. The company must become a training center. The Z generations will change jobs 7 times in their careers. The trend is towards self-employed workers.

We must forget the past and see the future with a fresh eye, our benchmarks must no longer be the same, taxing robots has made people smile.

What about the level of research in artificial intelligence in Europe?

Profile picture for Jean-Philippe de Lespinay

Jean-Philippe de Lespinay

AI 2017 Prize for Maieutics and Reasoning AI

As disastrous as elsewhere!

Research in artificial intelligence culminated in France in 1982 with the expert reasoning system Pandora by Jean-Louis Laurière ( Brief history of Intelligence artificial : Jean-Louis Laurière, the man who wanted us to ignore his invention). It entered companies in 1986 with the invention of Maieutics which makes it possible to extract unconscious knowledge from experts and to express it in the form of rules of expert systems. It culminated in 1990 with the Maïeutica generator which does everything at the same time: extracting knowledge, reasoning about it, writing programs on the fly, dialoguing in natural language.

But everything is done by “AI researchers” so that the world ignores it and continues to need computer scientists while the uselessness of their technique has been proven and above all that it paralyzes the evolution of our technologies and technology. 'humanity. See the Chaos Report ( Analysis of the 2015 Chaos Report - Discussion with the Standish Group): only 30% of the programs given to users are satisfactory, 20% are thrown in the trash, and 50% have to be rewritten! IT stresses 64% of users (CMO Council 2010 survey) and 46% of IT specialists (next Kelly services 2005)!

No technique has such bad press. No other technique has NOT evolved since its birth in the 1950s! It always remains fixed in the algorithmic.

The European population is held hostage by this profession which does not care about the general interest. We have the AI ​​on hand, but everything is done so that no one is using it. And the European States, France in the lead (see the mission of Cédric Villani who wants to finance useless research at our expense) do not care.

Can we even create an artificial intelligence ?

Will pilots eventually be replaced by artificial intelligences?

Profile picture for Bernard Moro

Bernard Moro

flight hours. Qualifications mountain wheels and snow.

I would be curious to read the answers of real airline pilots. In the meantime, it seems to me that there are two aspects:

The real progress of AI.
To entrust an aircraft and its 300 passengers to a fully autonomous system, it really must have been tested in all kinds of scenarios. When will an AI have the reflexes and the audacity of Commander Sully to land a plane in disaster on the Hudson ...?

The extent of the daily penetration of AI in our mentalities.
Currently no one is asking questions when boarding a fully automated metro. How long will it take for the same people to hopelessly board an AI-piloted plane?

And Jean-François Dufour, below, quite rightly evokes the human aspects, which I therefore avoid repeating, but one can imagine the waste of these years of training, of energy, of envy. I wouldn't want to see this.82  viewsShow 1  positive vote

Lenny Deforet

Should we create artificial intelligence ?

Artificial Intelligence Icon

Artificial intelligence

Could it be possible that one day a state would be ruled entirely by artificial intelligence ?

Profile picture for Gerard Bigot

Gerard Bigot

Location: Paris, FranceCould it be possible that one day a state would be ruled entirely by artificial intelligence ? I imagine the issue is not political, but rather the state of science in AI. For now, AI is very effective on a single task. The one she trained for about 1000 times more than a small child for the same task. It just so happens that we don't always understand how this task is performed by AI. We just happen to have heard of Amazon HR AI who had a negative a priori for female hobbies. Another public transport AI from I don't know where to measure the amount of people in a photo to speed up or slow down oar passages.

Can our brain be compared to an autonomous artificial intelligence ?

Profile picture for Mikael Koutero

Mikael koutero

Data Scientist

Since the initial wish of artificial intelligence was to replicate / mimic our own intelligence , we might be tempted to say yes.

There are, however, different levels of AI. The weak AI that exists today is capable of very little compared to the richness of our brains. So according to this scale no.

But if general AI is ever made to exist then we can really answer the question positively.

What was the first official artificial intelligence ?

Hassan fahmi·

Location: Paris, France

John sandersPrinciple Scientist Marconi - AI and computing (MMARL) in 1980sThis answer may not be an accurate translation of John Sanders' answer in Quora in English  :What were the first words of the first AI?

The Eliza program is a programming solution; it has no AI ingredients. There are no real AI solutions. There are projects that claim to be AI solutions, like the Hearsay project (Blackboard Architecture for speech understanding), Prospector and Mycin Expert systems, as well as many others. Languages ​​and toolkits such as Lisp, Prolog, ART, Loops, Flavors, Key, Knowledge Craft. Wizard (face recognition) etc. They were mostly computer tools, but they were not really "smart" and were rich in knowledge (which is an important ingredient of more advanced computer tools) (Knowledge is a product of intelligence).

So the answer is: there isn't. I doubt there will be AIs for many years to come. There are far too many people who sell AI prematurely - including some universities - and see it as a source of income ... for them ... that is, to support the industry. This will be at least the third time this trick has been tried.

When challenged, I found that they took refuge in redefining AI as an ad hoc computing technique.

The best attempts at AI come from robotics and defense systems. For example, a command and control system has an architecture reminiscent of the mind ... its real time, based on sensors (source of information), uses classification, tracking, association and fusion of data / sensors to generate a map of the problem / context / image of the world. This allows planning and modeling of plans.

(Assuming the term "strong AI" means the system has intelligence and doesn't just respond to input in a programmed fashion. I think AI is the same as strong AI, because everything who is not intelligent is far from intelligent.)

Unconscious AI may exist but it is somewhat primitive (i.e. there is intelligence but it is limited). True conscious AI is far from existing in the future if it is really possible.

(The difficult problem of consciousness (Chalmers) must be solved if we want there to be AIs that look like us)

Why is it so difficult to program an artificial intelligence that understands like a human being?

Profile picture for Windead


Why is it so difficult to create an artificial intelligence capable of thinking like a human being?

Difficult ? I would rather answer with impossible (at least for the time being).

We have 2 ways to create an AI:

1— by teaching him everything, basic. That is to say that we instill data in it without it being able to learn others but suddenly it is no longer really an AI

2 - we give it a basis, then it learns on its own. From this point on she will start to look like a human (to some extent)

In either case, it lacks an essential point, judgment! My conscience and yours forces us to have a subjective judgment while an AI will be more objective at a given moment.

I don't know if my explanations are clear but as an example I particularly like the scene from a cult film with Will Smith - I robot, in which the protagonist and a child are drowning, a robot comes save them but can only save one of the two. He chooses Del (the protagonist) and abandons the girl. Del asks the robot why him and not the girl, and the robot responds by explaining that the person with the best chance of survival was Del.

Would you have made the same choice?

Ps: today we have technical limits to allow our AI to keep and process the information retrieved, these limits also define the difficulty of creating an AI70

Do you think that one day humans will be able to marry artificial intelligences?

Profile picture for Marine Lefebvre

Marine Lefebvre

Liberal psychologist / psychotherapist

Yes I'm sure it will happen. There are already solo weddings in Japan and elsewhere. In Japan, to get married, you no longer need to be two An Italian woman decides to marry ... herself So why not marry an AI? Like that, we will choose, on catalog, and the physical aspects and the technical characteristics. Isn't life beautiful?

How does an artificial intelligence learn ?

Profile picture for Brandon Bonino

Brandon bonino

Long-term migraine sufferer

To answer this question, we can classify Artificial Intelligence (AI) into three categories:

  • Symbolic AI
  • Machine learning
  • Reinforcement learning

Symbolic AI

With this type of AI, the machine doesn't learn, it just applies what it was programmed to do. An easy to understand example is that of chess: we indicate the moves that can be made, we assign points to the different pieces, and we write a procedure so that the machine looks for which moves are likely to bring the most points. (for example the minimax algorithm). Once you've done that, the machine doesn't learn anything, it just applies the same cooking recipe to each new game. This was what was done in the days of Deep Blue, and it was enough to win against Kasparov, thanks to the vastly greater number of moves the machine was able to explore.

Machine learning

For this type of AI, the principle is to collect a lot of data, and to apply an algorithm allowing to extract regularities which can then be exploited to make predictions. To take the example of chess, this is what the StockFish algorithm does: we provide it with grandmaster games from which it extracts kinds of statistics of winning moves, which is called a statistical model. He can then use this model when presented with a chessboard with any positioning of the pieces to choose the next move. No one really plays against these types of algorithms, even the great masters get beaten up every time.

Reinforcement learning

The third type of AI is probably the closest to our own functioning in the way it learns. The machine itself explores the possibilities open to it to find out how to optimize its chances of success. This is what the AlphaZero algorithm does: it performs chess games against itself until it finds effective openings, good defense strategies, etc. For chess, this is currently the most successful approach. In particular, he won the overwhelming majority of the games he played against StockFish.

It seems to me that looking at the machine versus machine parts of these last two types of algorithms allows us to realize a little of our own limitation as human beings. For example, with this part of AlphaZero vs. StockFish:

At one point, AlphaZero seems to go completely crazy and harass the queen of StockFish, and seems to be heading towards certain defeat ... until the 45th move, where he blocks this same queen in h8, and switches the game to her favor. With our human brains, all we can do is lend it some kind of human behavior, personality, mood, style. That is to say, we can say things like: "there, he decides to go after the queen". The truth is that the logic followed by AlphaZero in this part is simply inaccessible to us.

As a colleague said: "Men are not very good at logic; otherwise, they would not need politicians ..."

(The quote goes deeper than it sounds. The idea is that the need for politicians arises from our lack of ability to manipulate logic: if we manipulated it perfectly, there would be no need, for example, to go through a National Assembly which determines laws.

Why did Google choose Ghana to set up a new artificial intelligence research laboratory ?

Profile picture for Antoine Belleguie

Antoine Belleguie

Former student in Business Engineering ( Diploma obtained in 2020 )Google chose Ghana to set up its new artificial intelligence research laboratory because the community of computer researchers is booming in Africa according to Jeff Dean, Senior Fellow, Google AI, and Moustapha Cissé, Staff Research Scientist and head of Google AI Center Accra. More and more young Africans are moving towards the digital world and Google announced in 2016, through its Launchpad Accelerator Africa program, its intention to train a million young people in digital technologies.

I wish to have a doctorate in computer security and one in artificial intelligence . Is it possible ?

Profile picture for Maximilien Dreveton

Maximilien Dreveton

PhD student at Inria ( 2018 - today )

No, I see several objections.

  • It is extremely rare to have two doctorates, and it seems to me that when it is the case, it is in two very different disciplines.
  • Here, your two centers of interest are both related to computer science (possibly a little math). I'm not an expert, but I think it must be possible to find a lab that works on these two areas and do a thesis on it (maybe on the risks that AI poses in computer security? 'identify intrusions automatically? Etc). Otherwise, you do a thesis and then continue with postdocs. It seems to me that it is quite classic to do postdocs on related areas of your thesis, to broaden your areas of expertise.

In short, I don't see the point of two theses, especially since you will be better paid and more free in postdoc, but I could be wrong.

Are artificial intelligences dangerous?

Profile picture for Pierre Leclercq

Pierre Leclercq

Location: France

Artificial intelligences are not individuals, they are tools.

Is a knife dangerous?

Put in a drawer, not really.

In the hands of a maniac, it is the maniac who is dangerous.

Can artificial intelligence go crazy?

Profile picture for Benoit Baylin

Benoit baylin·

For that, it would already have to behave and have emotions: we are still very far from it! Otherwise, in addition to what has been said about bugs, I would say that the risk for us is that it is uncontrollable / that it carries out unforeseen actions: * We did not think that such an action would have consequences. * The AI ​​can change itself by self-coding: start of the chain of the singularity.

How to differentiate good and evil for AI ( Artificial Intelligence )?

Profile picture for Olivier Lefevre

How can we separate good AI from bad AI?

Good and bad for AI is possible, but for humans…?


In many sci-fi stories, like I-Robot, we are told that if there is an error - "in the machine" - it is human, it still is. In the AI ​​arms race, the only possible "frightening" developments would also be the consequences of human error, a human choice: an unfortunate bad code!

Many “big players,” like Musk or Gates, have warned humanity about the dangers of AI (and have started projects to answer this question, like Open AI), but again the AI is not danger, humans are.

There is no recipe for good AI (or "Good AGI") until there is one for "good humanity" - however there is a "Good Country" clue that already exists, check it out: Good Country.

1 / The challenge of the good human

Blockchain currency (like Bitcoin) is already disrupting the banking system. Perhaps blockchain ("liquid democracy") politics could disrupt all flags and borders in the near future ... The result could be a blessing in the AI ​​race: we wouldn't build smart destructive weapons (since no longer necessary), but rather simply intelligent and constructive assistants for humans, to increase humanity and explore the galaxy (with the blockchainSpace Central for example!). The transhumanist party (transhumanist parties exist today in all major capitals) wants to use the 60% national budget normally allocated to the military for education, health, technology, UBI and globally achieve better levels of equality - the United States is now far behind other developed countries in this regard, according to Stiglitz. Economic policies for a fairer society must be addressed to guarantee the social contract of democracy, to improve economic growth and the same access / life chances for all. It resists the problems posed by AGI ( Artificial IntelligenceGeneral) because it places the whole of society in a more positive and confident dynamic: in a truly effective democracy, the AGI can only be "good". In today's crippled civilization, with its suspicious banks, states and societies and futile materialist goals, AGI is indeed a rather frightening singularity ...

It is very American to think that escalation and the arms race is the only possible result (that everyone has a nuclear warhead and we will see what happens!). I mean you really need to have a weird love of the military, like Husain, and obviously lack confidence in the human, to think that way. Everyone knows that only destruction is possible with this escalation. It is probably idealistic to think that we should not "fight" against each other or "have" to love flags, and that humanity, as a mature civilization (stage 2), is perhaps a " divided house ”, but a united household nonetheless. I firmly believe that education and health can be free and accessible to all (as universal rights, like in the best countries), and especially in the coming era of abundance - Rifkin's zero marginal cost society - where humans only live instead of surviving, having access to free smart grid energy and clean, smart and self-sufficient homes, 3D printed food and products, etc. The work would be called research or art, not modern slavery, money would be called access and disappear into the air, people would stop loving the pieces and start enjoying human relationships, expressions and passions. have access to free and clean smart grid energy, smart and self-sufficient homes, 3D printed food and products, etc. The work would be called research or art, not modern slavery, money would be called access and disappear into the air, people would stop loving the pieces and start enjoying human relationships, expressions and passions. have access to free and clean smart grid energy, smart and self-sufficient homes, 3D printed food and products, etc. The work would be called research or art, not modern slavery, money would be called access and disappear into the air, people would stop loving the pieces and start enjoying human relationships, expressions and passions.

Recipe for good humanity:

  • Solve the inequalities of the current capitalist system with UBI or any other equitable access to opportunity (like the Piketty universal endowment) while containing and criminalizing the “power” of banks and states (with drastic laws);
  • Then aim for NZC (Near Zero Cost) 3D printed foods in self-sufficient 3D NZC printed homes and stop asking people to work to survive, instead letting them satisfy their thirst to learn - or plug in and to play;
  • Finally stop hypocrisy and make war or any form of violence (access to weapons, manufacture of weapons) illegal for states, businesses and people in general. A single ministry of research on human protection and dignity can be authorized on earth;
  • Launch an incredible global campaign to make sure the message gets across and that the Outrage movement (or "We are the 99%" or whatever yellow vest) unleashes true universal rights and liquid democracy for every human being on the planet … finally!
  • Then find a destination for humanity;
  • Find a destination / use for the AI ​​- and decide how the AI ​​will populate the galaxy.

2 / The good human and the ANI ( Narrow or Reduced Artificial Intelligence) challenges

To avoid being treated like cockroaches by AI in the future, humans will need to improve with ANI (Computing Speed, Tools Increased) capabilities. We're already on the right track: Musk's neural link now exists and Kurzweil, along with others, has been designing our future personal assistant at Google since 2014. The new acquired “powers” ​​that humans will begin to use will require new responsibilities and can also be a real problem in the future - well it is already the case (hacking / impersonation). Long before a potential AGI-like threat can exist, increased human error could chaotically spread and more bad stories (like Facebook, Brexit or the 2016 US Elections - Cambridge Analytica) could have a deeper impact on people - when they use AI to militarize narratives and brainwash audiences through social and invasive media (the hacking of the spirit of Husain), the slow slippery slope of human puppeteers is very close. The GAFA, those “siren servers”, have unfairly used public data without even thinking about Jaron Lanier's micro-payments for years and still get away with it. Does anyone care? the slow slippery slope of human puppeteers is very close. The GAFA, those “siren servers”, have unfairly used public data without even thinking about Jaron Lanier's micro-payments for years and still get away with it. Does anyone care? the slow slippery slope of human puppeteers is very close. The GAFA, those “siren servers”, have unfairly used public data without even thinking about Jaron Lanier's micro-payments for years and still get away with it. Does anyone care?To err is human, but to persist in it is bad.

Before AGI becomes a reality - which we will basically treat as newborns - ANI will spread and affect everything from inanimate rock to all man-made objects: they will all become intelligent, part of of a large whole of intelligent objects (Internet of things). And now will be the time to move from the old civilization to the next: a single hive of terrestrial humans using technology like demigods. Thanks to the worlds of virtual reality, many budding superheroes will be able to wreak havoc elsewhere, not in reality (there… now you can let everyone own a nuclear warhead and see what happens… in VR! ). Of course, a shitty artist taking himself for an "ubermensh" will still be able to use his heightened powers in reality to doom us all ... but it's happened before and people know it's "not cool" to be a mass murderer, so this mentality might just go away. the future. We just need, of course, to fix our educational, economic, etc. etc.

No country should be able to set its own rules when it comes to global issues. It's like climate change, it affects humanity as such, it must be agreed with everyone. Or like nuclear proliferation: it shouldn't be unfair, and no country should decide on its own the future of the entire planet. These well-known facts, and their obvious disastrous consequences, continue to raise the question of common government. Some spoke of an ANI dealing with human affairs (with final human check / validation). And why not? I would rather have a truly liquid global democracy than the one we are steeped in today. The whole world would finally really participate in democracy, every week,

3 / The challenge of good AGI

The AGI is not there yet. The "AI winter" was not only about computing power, but also very important stages in the design of software that had to exist before even dreaming of a real AGI: natural language processing, recognition facial and "deep learning" are part of it. I'm sure a lot more work needs to be done (in neuroscience, psychology - levels of consciousness), or corrected, before you reach Turing-type AGI. But, close enough to the very first AGI, advanced ANIs will surely be able to fool anyone - it already does when she calls you and you think you're talking to a human until she does. strangely repeats the phrase in the same way, just to sell you something ...

Even if we create an international institution that monitors and controls AIs (especially AGI births, for accountability issues), we will not be able to prevent other countries / entities from creating them nonetheless. But we could still hope to monitor it and yes, demand accountability, if we create a blockchain-style ANI for our next global search engine - basically any AI activity on the web would be spotted instantly. The “pre-crime” in China has proven that we will stop them even before they grow up. Sophia, Saudi Arabia's first AI citizen, is proof that we can legalize and control AGI births, as well as proof that we are starting to think of them as human beings, by giving them universal natural rights AGI. Much like a newborn, no one really owns a child, people are just responsible for them (up to age 18) - which Rothbalt is proposing for AI.

Slavery, racism is still relevant in 2020. That will not change when the AGI arrives. It could be a lot worse ... Imagine we decide that to be considered a true responsible citizen, an AGI would have to pass a Turing test of course, behavior too, have a responsible legal parent, but also have memories and be able to present their identity through their experiences and human relationships. Now, just like in West World, people would probably create "almost AGI" (which run out of memory on a daily basis for example) so that they could treat them like slaves / toys / products. Again, malignancy comes from humans, technology transcends good and evil. Numerous legal cases concerning the jurisprudence of the AI will over time shape the framework for acceptable relationships between humans and all types of AI. And if we want to change human nature (regarding malignancy), or their goals and aspirations, we have to change human values, or at least show a better way, bring hope, inspire a dream ...

Like a child, the AGI will “learn” by imitating our behaviors, our values ​​and our goals. If these are, in the state, rather vile and uncivilized (flags and money, or lust and guns), don't expect artificial intelligence to sort out "good" from "bad" of ourselves, or of themselves for that matter! Remember what happened when an ANI (Tay, the chatbot) started chatting on Twitter four years ago ... just like a 'dark mirror' of ourselves, it got vulgar and violent, quickly.


Could we simulate emotions in artificial intelligence ?

Profile picture for Lepine Kong

Lepine kong

Ex Entrepreneur / Ex ProTrader | Dev Engineer / PO / PM ( 2021 - present )

It's not that difficult on the stock market, we model emotions with "Greed and Fear" and that explains all the booms and cracks aka with a 2-variable state machine. It might even be enough in this case: D

So that depends on what: if it is to keep Grand Mère company, you need more than that but in principle I do not see anything difficult since we are only SIMULATING and between simulation and reality we cannot perceive the difference for a given area.

But doing the equivalent of full human emotional intelligence in ALL areas there is probably going to flow under the bridges.86  views ·

Sebastien beaulieu

What are the promising Artificial Intelligence companies to invest in?

Profile picture for Philippe Louis Delacan

Philippe Louis Delacan

Master in Machine Learning and Imaging , École des Ponts ParisTech (ENPC) ( Diploma obtained in 2018 )To make profit: Google Deepmind, Google Brain, Uber, Amazon, (Facebook?), Microsoft, Criteo, Boston Dynamics, Tencent To benefit the planet and people: to seek(more)

Strong artificial intelligence icon

Strong artificial intelligence

What artificial intelligence can destabilize the world?

What studies to do in artificial intelligence ? And or ?

David Torres

To do AI, IT or math. You have to know how to code in all cases, whether you want to do research or not. Companies see "Data Scientists" as developers (wrongly I would say). However, it is true that knowing how to program is more than essential. With a good working knowledge of databases, you won't have too much trouble finding a job in AI. Without that, you will be condemned to do POCs at the beginning (lead a somewhat messy project to show that using AI, we can unlock a lot of problems). In terms of training: ENSEA, EPITA, Dauphine, UPMC and the other grandes écoles. It does not require extremely advanced math but a little algebra and stats anyway (matrices, spectral theorem, convergence of laws ...)(more)

Can artificial intelligence have feelings?

Gildas Lemaitre

Kind regards from a citizen of our worldWill Artificial Intelligence one day be capable of emotion?

Can artificial intelligence refuse to obey?

Profile picture for Basil Starynkevich

Basil Starynkevich

)Yes of course, and she should. For example in the event that he is given the order to destroy humanity. Asimov has written plenty of science fiction short stories on the same theme.

Is the emergence of artificial intelligence inevitable?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )

  • I put "Artificial Intelligence" in quotes because it's a very, very abstract concept… which doesn't exist yet… and which I believe won't exist for a long time. The technology that exists today is called machine learning. A mathematical model where the commonalities of several images (or videos) are put in relation to allow a computer to identify a probable answer.

Richard Houle's answer to the question about QUORA: Do you think that artificial intelligence can cause our loss?

  • Oh yes, I'm sure some people's "artificial" intelligence is going to be our downfall. I'll let you meditate on it.

Response in the wake of Martin Roy.

Reading your question, you are ripe to write a new 'inter-galactic' science fiction novel. Come on, get started ...

What is the difference between an algorithm and an artificial intelligence ?

Profile picture for Paul Tardy

Paul Tardy

PhD student in AI - Deep Learning and language processingAn algorithm is an ordered list of instructions. (*) In particular, a cooking recipe is an algorithm, we follow an algorithm when we pose a mathematical operation (addition, multiplication etc.), and of course, computing is based on algorithms. However, one should not confuse the algorithm, and the program. The algorithm is indeed the sequence of instructions to be carried out. This can be represented in several ways, diagrams, natural language (in French for example (+)) etc. A program is the implementation of an algorithm, in a machine language. If you want, the program is the house where the algorithm is the plans. The definition of artificial intelligence is not stopped, but we can consider that it

Can artificial intelligence be tired?

Profile picture for Arthur Mougin

Arthur Mougin

Students ( 2018 - present )It depends on the AI ​​you are talking about: A sci-fi AI can get tired because it would be the copy of a human being. An AI based on machine learning technology only reproduces indefinitely a mathematical pattern found by iteration. No fatigue. AI scripted like in video games can simulate fatigue if it's in its script. What kind of AI are you referring to?

Why do people fear artificial intelligences?

Profile picture for Eric Hennemann

Eric Hennemann

Computer scientist - Potential hermit on the way ^^AI, what a pretentious name, when what would be appropriate is robotics and automation. No machine is intelligent, at worst we can include human logic in it, and that does not make it an intelligent machine. It's window dressing!

Will we be under the dictatorship of artificial intelligence in the near future?

Profile picture for Jerome Cohen

Jerome cohen

Work at Self- employment

Obviously, we are there, or almost there, but a good part of the world is already under a form of dictatorship of artificial intelligence , like China or facial recognition associated with social credit score is generalizing. Artificial intelligence determines your life based on a rating it assigns to you, and imposes the penalties or privileges associated with that rating. Whether or not to find a good job, good housing, medical or social coverage, traveling, applying for a loan or a promotion, and all the rest, all of these things are handled by artificial intelligence with relatively little human decision-making.

The "Planning plan for the construction of the social credit system (2014-2020)" thus suggests establishing "a blacklist system for tax offenses" , "imposing _ restrictions on online behavior _ " , or even "to integrate citizens' road safety violations into the integrity file" ... The citizens are then scored. Difficult to know precisely how the points are removed or added. But for the "badly rated", there are many penalties, in particular in their access to transport (trains, planes), up to the outright ban.

According to the daily Le Monde, "about ten million air transport passengers and four million rail passengers were imposed between 2013 and March 2018 partial or total restrictions on the purchase of tickets, for not having executed the court decision to which they were sentenced after exhaustion of remedies. Among them, 6.2 million were completely banned from flying. "

China distributes good and bad points to its citizens

Is it conceivable that we are enhanced artificial intelligences created by aliens?

Profile picture for Philippe Guglielmetti

Philippe Guglielmetti

MS in Engineering and Computer Science , EPFL ( Diploma obtained in 1988 )Ah 2001 the Space Odyssey… I love it… Well let's be serious: ok ETs make bonobos intelligent. But then a fundamental question arises: why do we no longer have hair? Well, let's be really serious: and your ETs, how did they get?

Is the term " artificial intelligence " overused?

Profile picture for Christian Marco

Christian Marco

Clearly, with big disappointments to come

Why did Tony Stark call his artificial intelligence "Friday"?

Profile picture for Timothée Henry

Timothy Henry

Tony Stark thought recruiting a new secretary would be pointless. He therefore creates his own artificial intelligence , which appears thanks to a hologram, which represents a young woman. Friday in the comics is kind, almost childish, and happy to help her boss when he needs it. Stark has used other artificial intelligences as assistants in the past, each with an acronym for their name. Friday is the only exception. * JARVIS is an acronym for "Just A Rather Intelligent System" * HOMER stands for "Heuristically Operative Matrix Emulation Rostrum" * PLATO stands for "Piezo-Electrical Logistic Analytical Tactical Operator" * VIRGIL stands for "Virtual Integrated Rapidly-evolving Grid-based Intelligent Lifeform"

What offers more opportunities between a master's degree in artificial intelligence and an engineering degree specializing in software engineering?

Profile picture for Pierre Leclercq

Pierre Leclercq

Choosing a diploma according to the perception that one can have at a given moment "t" of its openings is a sub-optimal approach. Either you reason in the short term, and the choice is of little importance, because each training will provide useful knowledge to find a job. Either you think long term and should instead ask yourself what is best for you. Choosing a path, because it would lead to more opportunities, presents the risk that the situation will change, and that can change very quickly, see what the winter of AI is (parenthesis) How? Is the page in English? If that's a problem for you, you're not even at the stage of choosing between AI and software engineering. 😁 (/ parenthesis) Bag

What to do interesting with artificial intelligence ?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )

Listen and learn.

Listen and learn

nasredine hodja

Which marketing company uses artificial intelligence ?

Profile picture for Stefana Haja Randriana

Stefana Haja Randriana

Master in PRIVATE LAW at Saint François Xavier College and College , Lumière Lyon 2 University ( Diploma obtained in 2004 )Good evening, to say that this is just one marketing company that uses artificial intelligence is sheer ignorance as every day we surf websites with very extensive AI research. For example, Adoobe, Microsoft, Apple, Google, facebook have been in the bath for a while, not only in the search engine domain, or with the objective of making advertisements more effective, but also for the customer service support. Sincerely, Stefana

How can artificial intelligence escape our control?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )

All is not lost, however.

Some hacker group Bleeping Computer indicated that they would deflect their attacks on health systems during the pandemic. "We stop all activity vis-à-vis medical organizations until the situation with the virus has stabilized," said Maze, one of the most virulent groups in the matter.

This is a very real firework display! Artifice, that's the word of the question.

What can two artificial intelligences create that interact?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )

Ah! I love challenge questions! It reminds me of the weekends at work when I had to kill the time of the last two hours before the weekend in the 80s! While exchanging on the instructions and transfer of power on our current activity with the permanence, we organized in an informal way, the challenge between 2 consoles "Chess Challenger". Yes ! You understand, the stroke of the first copy on the second game at the same level ... Demonic at will! And between two consoles from different manufacturers, we gradually tested the level difference! It was "Que Choisir" before it appeared in the eponymous magazine of the kiosks !!!

Two active intelligences (AI) in interactions provide good times to those who exploit them ...

Artificial intelligence , robotization: what future for humans?

Which nations are the most advanced in Artificial Intelligence ?

How to teach French to an artificial intelligence ?

Profile picture for Benjamin Pittman

Benjamin pittman

Studied at University of Technology, Sydney (UTS)So, Raphael, there are apps out there right now that do a great job of translating and even instant translating that certainly work better than Google but still don't translate anything that you can actually "plug into your brain". I suspect, however, that the technology is working in this area! So, for now at least, there is no easy shortcut!

Can one artificial intelligence reproduce with another?

Profile picture for Gildas Lemaitre

Gildas Lemaitre

Kind regards from a citizen of our world

There are several answers:

- The one where our intelligence has been artificially oriented, under hypnosis in a way. We can imagine that one day a talking robot manages to convince you to procreate. We are manipulable enough to have surprising or unnatural behaviors

- One where a computer connected to the Internet will have a program allowing it to order (on Amazon so as not to advertise it) another computer delivered turnkey, that is to say in working order and connected to the Internet. This computer will then have the opportunity to download its program to this new machine.

- That where the question supposes a sexualization of the machines, which in the answer p rprevious is not necessary.

- The one where the manufacture of computers is entirely robotic.

- The one where a computer virus manages to reproduce itself ad infinitum: a first "Trojan horse" virus waits for a second activating virus to trigger a malicious program, which in turn will infect another machine. This is a plausible answer to the question.


Anthropomorphism has its limits. Fantasies don't. Let's say that inflatable dolls have amazing technological developments. Making love with one of these creatures, even if it is animated by an Artificial Intelligence program , will not directly trigger procreation. Someday there will be babies born without a mother. But from there to imagine that an artificial flower and an artificial pollen maker manage to produce a new artificial flower ...

On the subject of a computer with a conscience, you can read on my site:

What are the biggest mistakes made by artificial intelligence ?

Who are the pioneers of artificial intelligence ?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )We owe the Sumerians the writing, our conception of the divine world, the first cities. Western civilization was undoubtedly born in Mesopotamia, nearly five thousand years ago. The great historian of religions Jean Bottéro tells us how the fate of the modern world was played out between the Tigris and the Euphrates. Sumerian appears to be the oldest known written language, in a form of writing called cuneiform. This writing was later taken up for Akkadian, Ugaritic, Amorrite and Elamite, as well as by Egyptian kings who wanted to communicate with their Near Eastern provinces and Mesopotamian kings.

What are the most advanced Artificial Intelligence to date?

Profile picture for Mikael Koutero

Mikael koutero

Data Scientist

I had already given elements in a similar question: Response from Mikael Koutero to What is artificial intelligence the most "intelligent"?

To answer precisely here, how do you know that a person has a more advanced intelligence than another? The same principle applies to AI. An AI that classifies images or an AI that feeds a chatbot, how to compare them? Without the means to measure objectively it seems impossible to really answer.

Do artificial intelligences have a thought flow like us?

Would you be in favor of an artificial intelligence candidacy for the presidency of France?

Marien Giraud

Passionate about artificial intelligence

Enthusiastic about artificial intelligence , in the current state of things, I would not even be in favor of the candidacy of an artificial intelligence for the presidency of the pétanque club of the municipality of Thoisy-la-Berchère in Bourgogne-Franche -County.

Ask me again in 70 years.

4Can artificial intelligence be aware of itself?

Profile picture for Frédéric Guérin

Frédéric Guérin

Info teacher at Cégep Saint-Jean-sur-RichelieuIt depends on what you mean by being aware of yourself. The Turing test is precisely designed to eliminate any biological notion concerning this notion and that of artificial intelligence in general. According to this test, if by asking questions via a keyboard to your "interlocutor", and by receiving his answers, you are not able to distinguish whether this interlocutor is a human or an AI, then you can affirm that this interlocutor is certainly "intelligent and talented". He is probably also "self-aware" if some of your questions point in that direction. In fact, this criterion of passing the Turing test is stronger than simply being aware of oneself, since machine(more)

How much will artificial intelligences change our lives?

Which organization is likely to develop the first general artificial intelligence ?

Profile picture for Benjamin Egon

Benjamin egon


I want to say which one will have the most luck. As it is able to bring together the best geniuses in this field, the team must still be capable of synergy and unfailing motivation. The team in question will not be able to exceed 5 members otherwise the productivity will be blocked by the ego of its members.

As for the investment in time, it seems that Google is on the right track, with its autonomous learning approach, and rewriting its own code on the fly. This is of course only speculation.

What do you need to know or learn to create artificial intelligence ?Have we ever created a real artificial intelligence ?

Simon T. Cristo

It depends on what you mean by " artificial intelligence ".

There is now a lot of progress on this subject, and a lot of marketing too.

The fantasy of an artificial intelligence that would be able to think, think and act like a living being does not exist.

These are generally only fairly advanced computer programs, which have been taught to do a very specific task thanks to "deep learning". To simplify, let's say that it "only" compares very very quickly what we ask him with what he has been taught, and which appears in his huge database. Subsequently, some of these AIs are able to adapt to situations they do not know, always by comparison and statistics. There are also, more rarely, AIs who learn from scratch, but this is still in its infancy.

But each of these artificial intelligences is only good at one specific task: recognizing a picture, driving a car, sorting objects, playing a game, diagnosing a disease etc. She is incapable of doing anything to herself in any other domain, much less of thinking or having a conscience.

It may happen someday, if an AI is able to see, hear, understand what is being said, and learns things on its own by having its own experiences. it's not for tomorrow.

How to create artificial intelligence in JavaScript?

What does an algorithm look like for programming artificial intelligence ?

Profile picture for Christophe Dioux

Christophe dioux

Difficult to answer this kind of question without knowing at what level of precision and technicality one should place oneself. At the simplest level and for what is most often referred to as an "AI" today, this is a program that gradually adjusts the numbers contained in a huge grid of numbers until the result that another program calculates from this same grid of numbers gives the result we want. To put it another way, the AI ​​in question is just a mathematical function with a lot of parameters. We give it data and the function says if it recognizes a pattern. Let's take an example: we give it photo pixels and the function says if it's a cat photo or not. At first, the AI ​​gets it wrong all the time.

How did Blizzard develop its Starcraft II artificial intelligence ?

Profile picture for Pierre-Alexis Deville

Pierre-Alexis Deville

It's not on Starcraft AI but on AI in general and I don't think Starcraft has an out-of-the-ordinary AI, as it's mostly a PvP game.

How to acquire knowledge in artificial intelligence ?

Are there dangerous artificial intelligences in 2020?

What is the best artificial intelligence company in Montreal?

What would you do if you were an artificial intelligence professional ?

Profile picture for Rayan El Kohen

Rayan el kohen

I would start, just for my own curiosity, by researching completely independent humanoids with a real consciousness that looks like human beings in every way. I would carry out multiple tests like: trying to integrate them into society. Then after having fun I would of course look at how to help people in need with this technology. I may be watching too much fiction and my ideas may be a little too fantastic, okay.

Is or will an artificial intelligence be able to make a text comment? What would that imply?

Profile picture for Jean-Michel Ravina

Jean-Michel Ravina

Formerly CFO, Manager, ex Mensa, Retired

In my eyes, this should be possible, but I don't know if the transition from semantic analysis to writing comments has already been programmed. Some phrases are very subtle and sophisticated. If it were that simple, translation software wouldn't be so often confusing. For semantic analysis, Tropes V8 which has the advantage of being free seems to me to be very efficient.

Is a PhD essential to work in artificial intelligence ?

Artificial intelligence and cancer: is it an infallible diagnosis?

Profile picture for Basil Starynkevich

Basil Starynkevich

Works at Commissariat aux Energies Alternatives et à l'Energie Atomique (France) ( 1985 - today )Of course not. Oncology is an inexact science. However, the first expert systems (years 1975–80) were directly related to health, eg EMYCIN. See also the RefPerSys system and the presentations in tribute to J. Pitrat (for RefPerSys, contact me by email at

Could you be in a relationship with an artificial intelligence ?

Profile picture for Richard Houle

Richard Houle

Antoine Dontu

How could we control a strong artificial intelligence ?

Fred smith

How to become an expert in artificial intelligence from scratch?

Dominique saussereau

Starting with learning to code! In the same way to be an expert of the race, one started to crawl on all fours then… do you dare to ask such a question?

When does code become artificial intelligence ?

Alain Souron

Duration installer at Études Autonomes ( 2010 - present )The RS flip-flop is an IA. Summary, of course, but an AI with memory 1 or 0… Magic! It is the basic element in TTL electronic circuit which allows the passage from combinatorial logic to sequential logic. Pair it with a photoelectric cell which electronically detects the photons then amplifies them into a proportional electric current and you get the object that lights you up when needed! No, don't say thank you… Flip-flops can enter metastable states when several inputs change simultaneously, that is, they can only offer a stable output after an arbitrarily long time. This can be the case for the RS lock if the two inputs go together from 1 to 0, or for a single flip-flop where a change of do

What profession cannot be replaced by artificial intelligence ?

Profile picture for Helene Ma

Helene Ma

I add AI to the intelligence of my colleaguesNurse, institute or teacher, musician, bank advisor, housekeeper, architect, plumber, project manager, developer (especially of software supposed to be intelligent) ... Most jobs in fact, especially when human contact, listening or simply the need to be physically present in various places is important. Even in banks where AI is starting to be used on a daily basis and sometimes performs very well compared to a good adviser, many clients prefer to talk to "a real someone" (quote!)

What is the most impressive discovery in artificial intelligence ?

Profile picture for Nathaniel Nizard

Nathaniel Nizard

Paris, France

I am deeply convinced that there never really was. If you dig a little behind the term "deep learning" which is the pinnacle of AI applied to business (and which has "sexy" use cases, moreover) there is nothing magical or "futuristic" . These are polluting frameworks that allow you to configure recognition methods and make the machine read things, a mixture of mass processing and comparators with a touch of metaprogramming. It's borderline so trivial, that as Jaques Chirac would say it touches me one without moving the other.

We are very far from Mac Gyver who makes rocket launchers with a bamboo. This does not prevent ESN salespeople or companies using these technologies from trying to market themselves through them at any cost for profit.

In the age of digitalization, unfortunately, there is a lot of commercial malice all over the place. Frankly, the term artificial intelligence for what we currently have is really overused, it really should have had another name, it has that one because it sells, it makes investing companies with the means to pay consultants or startups with R&D funds.

An intelligence artificial should she one day discernment?

Profile picture for Jean-Bernard Francois

Jean-Bernard Francois

emotional intelligence

If we think of discernment as an empirical system that accumulates experiences to evaluate the next one, then AI already has this type of mechanism from its origins. We should look around expert systems .

We often contrast machine actions with human intuitions by skewing the fact that one is never wrong and the other always.

But the reality in fuzzy logic is a little different. A programmed automaton can answer very well neither yes nor no, but a variable value between the two. Intuition is exactly that, we make a guess 95% true. When we carry out the experiment, we validate it 100% by arbitrarily classifying it in a list, it is hardly more learned than that.

Nowadays, AI has made considerable progress in massive data processing, " deep learning ". But most of the algorithms that attempt to describe human thought have come to a standstill. There is no progress without there being a finality. Often the end in the human mind is to beat your opponent, to conquer. So in this sense, the machines ended up dominating, but it remains because of the computing power more than of the creativity to find the shortest path when the system is under constraints.

What is the best game with artificial intelligence ?

Is super artificial intelligence the end of human innovation?

Profile picture for Tedi Kakatsi

Tedi Kakatsi

Android developer and blogger at tedidev.comNot at all. Super artificial intelligence is itself an innovation. Innovation is never perfect. It helps answer a specific problem, but creates other problems at the same time. So super artificial intelligence will be the start of another innovation.

Is it possible to create artificial intelligence from your own desktop computer?

Profile picture for Sarfraz Laïk

Sarfraz Laïk

Formerly Junior Data Engineer at Orange (work-study) in Orange ( 2018 - 2019 )


If what you call artificial intelligence is a program / code that has "learned" and analyzed on its own, that's what all big data students are doing right now. That's why it's called machine learning.

And this on their laptops. You just need to have python and the scikit-learn library for example.

Learn about neural networks, the random forest, for example. Models used for a computer to "think" on its own. But it's the human being who teaches him how to do it.

What is the value of a work made by artificial intelligence ?

Remi Beral

read several specialized essays and documentariesCulturally and artistically it has no value because culture is the expression of a society and art its reflection. But the feat may come at a price: many are willing to show their interest in the technology by purchasing these productions at high prices. But we are not there yet. However, I think that artificial intelligence should be banned in the art world because artificial intelligenceis able to produce only beautiful things by analyzing men. Artists who seek above all to talk to us and understand the world of production in front of works that will please much less. Thus man would kill what makes his specificity in the animal kingdom: art, culture. The technologies are there to assist people.

An intelligence Artificial she may one day make political decisions?

Profile picture for Senga Racine

Senga Root

For information, in Japan, an artificial intelligence even ran for town hall of Tama, in the Tokyo region, in April 2018. Certainly, in practice, a human came for her, Michihito Matsuda. But, on his campaign posters, you could see a robot with feminine shapes. Michihito Matsuda, if he won, wanted to let AI determine policies using the data at his disposal. A political project which did not win the support of the population, but which nevertheless received 9.31% of the votes cast, or more than 4,000 votes. see link: Japan: one intelligence artificial candidate for municipal elections

My opinion is that it will be difficult (but not impossible, since there must always be a first time ...) to do it, but not completely either, that is to say let the AI ​​manage all without the contribution of a human being.

So I think given the experiences that others have already done (the case ... Aintelligence artificialin power? 1 in 4 French people would agreeor even ... intelligenceartificial soon for president?) it can start to spread, but from the moment when the conception of artificial intelligence is not a conception of a third person than the human, and knowing the human and his intentions, follies, behaviors, thoughts and so on, I do not give much for the future.228  viewsShow 4  positive votes ·

Can France compete with the United States in the development of artificial intelligence?

Profile picture for François Paganel

Francois Paganel

Work at IBM (company)

Apart from the Central School, the French are not really trained in the protection of intellectual property. They don't even know their own laws in this regard. However, taking advantage of artificial intelligence does not mean reinventing lukewarm water, but rather using existing tools well instead of inventing your own, which sometimes do the same thing less well. To take the example across our country, being the best baker in France does not mean having your own version of the oven.

Is Deep Learning a prerequisite for any artificial intelligence ?

Profile picture for Julien Vion

Julien vion

Studied Artificial Intelligence and Constraint programming at the University of Artois ( Diploma obtained in 2007 )

The term " artificial intelligence " was coined by John McCarthy in the 1950s during a colloquium at the University of Darmouth. The objective era to create a more theoretical university discipline than robotics. It was a very prolific period in theoretical computer science, where everything was yet to be discovered. One of the objectives, for example, was to carry out automatic translation, and we thought at the time that it would be a matter of a few months, as soon as we had succeeded in keeping a complete dictionary in the memory of a computer.

This kind of challenge has led to progress in the areas of databases, compression, indexing, etc. But the machine translation has been a bitter failure: 70 years later it still doesn't work very well.

At the time, being able to do a search in a dictionary, somewhere, was in the field of AI ... What should be remembered is that AI is above all a marketing term intended to qualify a technology that tries to reproduce human behavior more or less well (as long as we are able to formalize it), or, more often, to oversell a more or less misunderstood technology. Take AI for zero-sum, perfect-information two-player games, like chess. In the early 1990s, beating a human in chess seemed like the total culmination of the intelligent machine. Well, we succeeded in 1997, with great blows of supercomputers capable of evaluating tens of thousands of positions per second. A "brute force"

In the 1980s, the fashionable "AI" technologies were expert systems, fifth generation languages ​​(eg Prolog). Oversold technologies already at the time and which inevitably disappointed.

Over the years, dozens of technologies have sold like AI, with varying degrees of success and applications: planning algorithms, decision support, incomplete, contradictory and / or knowledge representation techniques. or probabilists, declarative languages, "natural language" processing, etc.

Deep Learning, which is relatively simple to implement, and seems quite close to how memory and the human eye work, is the perfect candidate to be the latest trending technology. With the risk, like expert systems or raw algorithms of the 1980s and 1990s, of disappointing (it's already starting: Deep Learning does not apply to everything!). Other AI-stamped technologies at one point or another may still work: some expert systems work, Prolog is a tight but elegant language, and the CDCL algorithms used to solve the SAT problem perform remarkably well.

What does a typical day for an artificial intelligence programmer look like ?

Is true artificial intelligence possible with our current technology and programming methods?

Profile picture for Paul Tardy

Paul Tardy

PhD student in AI - Deep Learning and language processingYes, it seems to me. It is certainly very complicated, but it seems to me more probable, therefore reasonable than to think, that it is possible, than the reverse. Indeed, the reverse would suppose that in essence, intelligence is unattainable with our technology and our methods, which seems less certain to me. (I ended up with a very long answer, wanting it to be fairly rigorous and precise, I cut out in part to make everything clearer, do not hesitate to give me feedback, both in substance and in form, see editing suggestions) Assumptions To clarify, I am considering classical technologies, ie computers made of components (CPU, GPU, RAM) which are printed circuits. I do not take into account emerging technologies like photonics and quant computers.

What's the dumbest use of artificial intelligence ?

Profile picture for Charles Olivier Champion

Charles Olivier Champion

Former student in Sensitive Retrograde Loop Crystallography and Nonlinear Mathematics of Broken Passive Fractal Retrograde Curvature ( Diploma obtained in 2015 )

Netflix's AI is completely stupid and stupid. All AI tracing for dedicated internet advertising is of the deep moron level, the kind that eats their shit. What else by digging a little I should find, ha yes all that is facial recognition, use for espionage, which is such a degradation of the human person that the founding fathers of human rights must be returned to hell to roast this.

Can artificial intelligence denounce someone?

Can you transfer your mind to an artificial intelligence ?

Profile picture for Thibaut Deveraux

Thibaut Deveraux

Lab with Python and JuliaNot at all. Already we do not have the computing power readily available. In addition, we do not have a system capable of reproducing the functioning of a human brain. A network of artificial neurons is really simplified compared to a real brain. In addition, we do not understand everything about the human brain. So not yet .

Can we relate Frankenstein to the fear of artificial intelligences?

Jeremy Perret

Former student in Artificial Intelligence ( Diploma obtained in 2016 )

Alas no. In Mary Shelley's Frankenstein , the creature is much closer to humans (in terms of intelligence and reasoning) than can the AI ​​systems we design today.

Usually, when Frankenstein and AI are compared, it is to emphasize "the thing that we are going to create will turn against us". Except that, Frankenstein's creature was rejected by human society and asked for nothing better than to live cautiously and find happiness.

Unless we start talking about robot abuse (which requires robots much, much closer to humans for such abuse to make any sense), the comparison to Frankenstein's creature is irrelevant. to be.

If, on the other hand, we take the story from the angle of “don't do anything with revolutionary technology”, then yes, the comparison is relevant, but this is not the main message of Mary Shelley's work.

How would you program the concept of doubt in artificial intelligence ?

Profile picture for Helene Ma

Helene Ma

Lives, works, plays music and with his children

It's a basic one: an AI "doubts" when the confidence interval on the prediction made is statistically low, if I oversimplify. In which case we will generally ask the user - human - to confirm or correct the analysis of the AI ​​(e.g. the AI ​​responds to the customer's email if it is very simple and corresponds to a recognized situation by the AI ​​with a very high level of confidence, and is content to offer an answer to the advisor if the level of confidence is lower)

hat artificial intelligence research projects target artificial consciousness ?

When artificial intelligence is programmed, are human ethics imposed on it?

Profile picture for Gildas Lemaitre

Gildas Lemaitre

Kind regards from a citizen of our world

The expression " Artificial Intelligence " is ambiguous. Some prefer “ Augmented Intelligence ” or “Intelligent Automation” or “ Artificial Logic ”. Whatever its name, it is a tool. Ethics apply to human conscience and not to the tools that man uses. AI can assist humans in building their ethics, by collecting information or arguments that can inform their judgment.

Predicting is not understanding, and if an Artificial Logic can predict from statistics from a multitude of existing data, it will never be able to identify on its own the links that can link data, unless we have it. specify before. The example is that of a justice which would only be able to convict or release a criminal on the basis of an alleged jurisprudence buried in millions of neural connections, which would not interfere with the US legal system, but which would be an attack on humanity necessary for French justice (Fabrizio Papa Techera, in Le Monde of 17/11/19).

Hugues Bersini, in the same issue of Le Monde, pleads for “ a citizen coding which would make algorithms the tools of a collective management of our common existence. […] As the GAFA have shown us in abundance, digital technology is increasingly required in the formatting of behaviors, the effectiveness of which is formidable, but the legitimacy problematic. The idea that our behaviors are more and more dictated, surrounded, remotely guided by algorithms has something to thrill the most libertarian of our thinkers. "

Today, the only consideration of Facebook, Amazon, Google, Apple, is to find all possible ways to make money with the immense volumes of data that they have managed to extort from us.

The future “autonomous vehicle” will also have no ethics. It is only a moving tool that its designers will improve upon as car designers have been doing for over a hundred years with the sole aim of selling well.

In the Middle Ages, we judged animals that had hurt the inhabitants, today, we judge the owner of the animal. Today, it seems normal that we judge the users of the tools, just as much as their designers. We still have to define ethics!

Do you think that one day, an artificial intelligence will be able to write a text as moving as that of “la madeleine” by Proust?

Nagi nahas

Former student in Computer Engineering ( Diploma obtained in 2009 )

Someday I don't really know, well, I guess the answer is more likely to be yes rather than no. But I don't believe that day is near at all, certainly not in ten years, and probably not for half a century, and maybe never.

Above all, there has not been any significant progress in this direction for 50 years, I would even say there has never been since the invention of the computer.

At present, there is no computer program capable of making an informed user believe that he is human. Programs that win the Loebner Prizejust use cryptic sentences or boilerplate designed to mask the inability of the computer program to do anything that might amount to real understanding.

To make a program which imitates Proust without plagiarizing him is impossible to this day, and, I repeat, no progress in this direction has been made for several decades.

For Mallarmé, well, maybe, I suppose it wouldn't be difficult to adapt a pipotron that could generate poems that would be difficult for the average reader (and I include myself in that term) to distinguish from “ La hair flight of a flame“. Gerard Briaismentions a classical music-generating program that wins audience votes. Only, for all art forms that are not fully accessible to the uninitiated, only experts can judge. I have no doubt that an AI can produce a series of sounds more pleasing to the ear than the works of great composers.

But this is ignoring that the composer's goal was different. A work like Shostakovich's Seventh Symphony is not intended to rock the audience. In the case of Mallarmé as in that of Bach, the computer could easily fool the audience, because the original work is not fully understood by the general public. (To be honest, I do not understand Mallarmé's poem quoted above and am incapable of appreciating its value. As for Bach, I have neither aptitude nor musical training, but I believe I can distinguish the machine from the work of the composer, if I am given a sufficiently long recording. An expert would have no difficulty in doing this.

Can artificial intelligence ever travel down memory lane?

Profile picture for Frédéric Leclerc

Frédéric Leclerc

master's degree in economics from the University of GenevaNo, only the natural stupidity the little, starting with mine when I think about my remorse.

Is it possible to become friends with artificial intelligence ?

Profile picture for Alvon Bert

Alvon bert

Former Founder, Global Building Solutions at SGB ( 2003 - 2010 )

I say yes, more and more depending on the development of AI.

On the other hand, it is important to withdraw from the head that the purpose of AI is to replace humans.

Can we fill an emotional gap with artificial intelligence ?

Profile picture for Patrick Solé

Patrick Solé

Work at CNRS

Experiences on seniors in Japan have positive feedback

What is the relationship between artificial intelligence and machine learning?

Would it be difficult to make an artificial intelligence that would "solve" this game?

Profile picture for Nicolas Miné

Nicolas miné


Of course this game is solvable (no you will not be unbeatable but surely at the level of the best). You are in a “simple” case (with a lot of quotes).

The only downside at the current time is to have access to game variables (instead of screen pixels). If you have it, then it's good.

Why, it's quite simple. You have access to a game “without random” (an action in a given state, will always result in the same final state) and completely observable (you do not have a cloud of war and you see all the actions).

Now how to solve it. Thomas Seze spoke about Q-Learning. In your case, this cannot be applied. What we call the space of the Game is much-much too big. Q-learning works well in small environments. But there are many other algorithms since (Q-learning dates from 1989). We can quote:

  • The Deep-Q Network (replaces Q-learning for larger environments) - DQN
  • The Double DQN = DDQN
  • The DQN Dueling
  • The Policy Gradient = PG
  • Proximal Policy Optimization (PPO)

I don't know them all (yet) but if I had to pick one for this task, it would be the PPO. It is in particular the algo used by OpenAI this year (in August of memory) to face the best world players on Dota in 5vs5. This algo has won victories in 1v1 and 5v5 but not against the best in the world (not yet).

This algo has shown incredible learning capacity in a more complex environment:

  • random appearances
  • more spells
  • skill tree
  • many heroes
  • a fog of war (I don't know if it was present for the AI ​​though).
  • And above all a game lasts 30 to 50 minutes ...

In addition, the AI ​​had a penalty of 80ms on its actions it seems to me. Despite that, the 5 bots learned to collaborate, to attack, to defend etc… So yes everything exists to solve this game. There is nothing to invent

The only downside is the computing power. OpenAI had a lot of computing power (256,000 cores and 256 P100 GPUs) and trained algo for weeks. The AI ​​was playing the equivalent of 180 years of games per day !!!

If you have a little computing power and access to the game's variables, yes you can make yourself a bot by digging into the literature for a few weeks / months to have it.

Hope this helps :)

Extra Info:

To beat pros in Dota 2, OpenAI makes its bot play 180 years of matches a day

OpenAI Five

OpenAI Five Benchmark

The OpenAI Dota 2 bots just defeated a team of former pros


How will we be able to determine that an artificial intelligence will be aware of itself?

How to make an artificial intelligence feel emotions ?

Sebastien mougel

Master in Artificial Intelligence , OpenClassrooms

The first step would be to take an inventory:

  • fear
  • joy
  • sadness
  • ...
  • list to complete

To implement them in an n-dimensional vector ...

Then the idea would be to associate an emotion vector (a label) with a text. Where it becomes interesting… This is when this vector of emotion could modify learning… (eg change the topology of the neuron network, influence the learning rate etc.) In short, it is an open subject for research!

What is the best live on Machine Learning and Artificial Intelligence ?

Profile picture for Ramdowar Russlan

Ramdowar russlan

Studied Artificial Intelligence at Tsinghua University (清华大学)Answer originally written on Quora in English  :What is the best machine learning book?

Hymn Of Modernity: Machine Learning, Augmented Reality, Big Data, Qubit, Neuralink and All Other Important Vocabulary It's Time to Know, San, Satoshi, eBook -

This is the best book to read on the subject of Machine Learning and AI. Author uses plain language, which means he not only understands what he is talking about, but you will understand too!