Introduction

So thank you, Frank, for inviting me. What I want to talk tonight about is what is beyond AI.

So first of all, I apologize for my French accent. I have a strong one, so don't judge me too much with this.

So, also what I wanted to show you, I'm going to start with three objects. And what I want you guys to think about is what do they have in common.

So we'll start with a picture of blue cheese. We are in France. Why not start a presentation with cheese?

So that's the first one. You have a piece of Kevlar. And finally, the molecule of penicillin. So does any of you guys find what all the three things have in common?

Something important

Some details

A network. Like a network of things that is . The imagery is similar.

Also, I could have added the microwave because it's well known for that, the microwave, but it was too obvious. You already find it, but that's the point.

All the three things were discovered by mistake or by chance or by luck. And what I want to emphasize with that is that us, humans,

The way we develop the scientific method is through errors and failure. We have a thesis, we are starting from a point, then we are doing a lot of experience, fail, failure, and then we see if we are correct, if it's right.

Some more details

My point with that is to emphasize what's the definition of human intelligence, the way we are creating value, truth, and ultimately how we handle innovation, how we are going through that. And the point is when you compare it with AI, it's really different.

How did we construct AI? the artificial intelligence.

Basically, if you compare it to human behavior, it's quite different.

Because we are building on failure, AI is constructed to do zero mistake. So therefore, you see, if you compare it, two different ways to behave.

Case Studies

And now, what I want to say is how how we can define intelligence.

Jane X Case

And I want to show you some cases. Here, the Jane X case, where all our perceptions are changed.

Does anybody know this picture? So this picture is one of the first pictures to win a photography contest online on Instagram.

No suspense, this is an AI-generated image. 1It sparkled a big controversy because it won in front of a jury of expert photographers. that validated and said, oh yes, that's a great photo.

Now if you analyze deeply now that you know that's an AI generated, there's a lot of stuff that is not working on this image, for example. We don't see it really nice, but here you see there is two sort of zone. And here there's nothing. There's strange light. You have few details.

Now that you know, you can see in the picture, there are some elements that are not really realistic. But the point is that it marked the history with the fact that someone has been able to fool some professional expert jury with an AI-generated image.

For some, it says that this picture has passed a kind of Turing test, where humans are not able to say whether it's photography or AI-generated. So that's one of the marks in the AI-generated picture history.

Jason Allen Case

There's another case. Does anybody know this one? I think this one is more famous, right?

This one also was made by Jason Allen. This case is quite interesting. It's representing a space opera scenery kind of, but it won a fine art contest.

It's made with AI, of course. But the whole thing is that the artist didn't try to pretend that it was made by human means. But it brings a whole question. Is it cheating to use AI to participate to fine arts contests?

I don't have the answer. I'm just asking questions and presenting you facts. This one won. And it brings that question.

I think it's fair to ask. So that's interesting.

Susie Dougherty Case

This one, Susie Dougherty. It's also a really interesting case.

That image, basically, It was presented at a photography contest.

And Susie has been canceled. Well, the jury decided that, no, we are not receiving AI picture because it's photography. But the point is that this is a real photo.

And you have that kind of look on the people because she used mannequin. And that's where it shifts the whole perception we have, you see, because once again, the Turing test is validated. We are not able to say whether it's AI generated or done with photons and photography.

And she made it on purpose. And also, by that time, we were on the Mid-Journey 5.1. And that's true that when it was generating several characters, it had a tendency to have that mannequin looks lifeless characters that mannequin can simulate.

And when there was an interview of Suzy, really interesting, people were asking, were you not too ashamed being canceled of that contest and everything? She just answered, that is strange because she felt like the whole industry felt into a paranoia. where they are like, OK, this is AI-made, this is not. But nobody is really able to say.

And she has metadata. When you're taking picture, you have metadata attached to the photography so she can prove that it's a real photo. So that's also a shift on perception, something that most people don't really realize.

Boris L. Dachshund Case

And now you have this really interesting example of Boris L. Dachshund. German artist, where he won that great photography contest with an AI-generated photography. And he refused the price to make a statement.

His point was, OK, the industry really need to decide whether you can participate with AI picture on photography contest or not. Or shall we create like a special kind of contest for AI generated picture? Again, I don't have answers. I'm just asking question.

But yeah, he fooled all the industry also. It was like, it's a really well-known contest in photography. The jury is made of real professional that have been making, they have been making photography for their old carriers and they fell for it again.

Myle Ashtray Case

And lastly, but not the least, you have Myle Ashtray, who won an AI contest, image contest, with a real photograph. That's a real photograph of a pink flamingo. It is lacking the head, because he's scratching himself with the backward. So it looks kind of AI-generated, some kind of hallucination. But he won.

And then, again, our perceptions are all over the place. So shall we be able to participate in this kind of contest, AI contest, or not? I'm asking you. So yeah.

We are talking a lot of AI bias, AI hallucination, AI errors. You all know about it. You've encountered it if you're using AI. But we are having, us humans, a real problem to even evaluate the work of as a human using AI or not, it's really difficult.

The whole point is our perception are totally shifted with AI and we need to, to my perception I think, we need to settle some new definition of stuff.

AI Perception Shift

So now let's talk about the Now, the elephant in the room.

Intelligence vs Stupidity

We are talking a lot of intelligence, artificial intelligence, of course, human intelligence. But if there is intelligence, there is also stupidity.

And that's what I want to talk about, about AI BS.

AI Errors and Hallucinations

And if you're using AI, you must know that it's doing a lot of errors, a lot of glitch, a lot of hallucination. But what can we say about it? Most of it is made of how it's programmed by human. It's coming from human. Of course, it's not coming from the sky.

Also, one of the key points of the way the AI is creating value is more on this way to... How should I say? It's what is bring... in terms of conception, in terms of imagination.

It means for me, as an AI user, you can have a real interest in having hallucination. For example, when you're using ChatGPT for search or for something like that, you want it to give you the truth. For example, if you're asking ChatGPT about winter sports events in Dubai, You want it to say, no, it's not existing, this is not real.

For previous version of it, it could answer you. That's something I've tried. It was answering, yeah, and it was creating a story from the top of its head.

Creative Use of AI

Now, what I like in those kind of stuff, hallucinating, I think it's an advantage for everything that is creative when you're using it for picture. because you all know Google image generator had had some problem when you were asking it, for example, can you represent like a white president of the USA? And it wasn't able to do so because it was programmed in such a way that they were trying to push on minorities. So that was an issue, right, that he was hallucinating.

But me, as a creative person, I like to push the boundaries of the AI to create some happy incidents. And that's, to me, that's when it gets really interesting is to see when you have algorithm that is able to go into deep database and try to create like a melting pot of everything. And you, with your human eyes, you create a meaning, a symbolic talk, because the AI don't have that kind of talk.

I'm sure most of you guys have tried some AI image generator, and if you're talking into kind of concept in symbolism, it's not really good. You have, as the whole prompt engineering approach is, you have to use specific token, specific structure in your prompt to have some specific results. And the point is that if you're able to create nice prompt and have some result, it will be you, as a human, that's going to create the symbolism.

Conclusion

So to me, the whole thing that brings me to the last aspect of it, not this one, is a complementarity program between AI and human. The whole thing is that it's a relationship between human and AI.

Maybe one day we will reach to the singularity, maybe. But for now, on this path, it's a relationship that we are creating with another type of intelligence, an intelligence that is working clearly differently from us, that is not programmed the same way.

And also, like in every relationship, we need to determine the terms on how we want our position in that relationship and not wait for the AI to determine it because otherwise it's already done.

Thank you everybody.