AI cannot imagine something new and original that can have impact years later: Laurent Daudet

Is it a dystopian fantasy? A comic book thriller? Or is the message here more urgent? Written in the decade of ChatGPT and digital assistants in a collaborative effort by scientist Laurent Daudet and graphic artist Appupen whose given name is George Mathen, Dream Machine: AI and the Real World (Westland, 2024) is, in equal parts, all three. Should AI be regulated? Well, the threat of AI-driven job loss and corporate takeover of the world is more real than was anticipated, the creators tell Sucheta Dasgupta in this interview.

How did this book come about?

Appupen: I was in France for an artists’ residency and met Laurent through a common friend. I had not conceived the idea for the book at the time. It was his perspective on the subject which prodded me to know more and more, and finally propose to do something together in a comic format. He told me about batteries and how much resources was going into this research and so on; and opened me up to a lot of things that my tech friends would really not talk about.

Laurent: It was pure chance. But then you have these two persons meeting with complementary knowledge. The original project was a short explainer, but grew into a novel.

Is it your first graphic novel?

Laurent: For me, yes, definitely.

Appupen: I have published six graphic novels. I have been publishing from 2009. I write and I draw. Usually, I want to make my readers aware about what’s going on with what knowledge I have. And here I have a lot more, especially with Laurent coming in.

What was the first book?

Appupen: My first book was called Moonward and it was published by Blaft. It was a complete fantasy but drew from the happenings around the Bhopal gas tragedy.

So Laurent, what has been your background in AI?

Laurent: I started my career at the university being a professor in physics and information theory. In 2016, I co-founded this startup called LightOn where I am presently employed, where we do large language models and provide alternatives to ChatGPT for the enterprise market.

Is there any cutting-edge research on making sentient AI and, if so, is the self-interest of the sentient AI bound to clash with that of the humans?

Laurent: We are still very far from any type of sentient AI. This is a fantasy and part of the Big Tech narrative that it is spreading to obfuscate some of the real challenges that we have like job losses and copyright issues. Sentient AI is something you find in science fiction movies but not something you find in the labs. I think, on a scale of zero to 10 of sentience in AI we are somewhere between level two and three.

Appupen: Laurent says AI is stupid. That is when I got interested. They are just algorithms. They have no notion of truth or values.

On that note, I have two questions. Can you explain the term, artificial general intelligence? And, since as you say, AI has no judgement of value and as I would suspect, having no emotion, no clue towards recognising what is true, good and beautiful, why is it that entrepreneurs are launching publishing startups that use it to play editor, select and finalise from submissions which novel manuscript to publish? Isn’t it a no-brainer not to do so? And in that case, how real is the threat to job loss, at least in the world of the arts?

Appupen: AI does not create, it reproduces; and that is one thing to remember, the ability to teach itself to reproduce without external input, on the basis of memory which is data already fed into it, being what is called artificial general intelligence. So it has no intuition while it is far more efficient than the human brain. It is super-fast and there is no error. Now, the machine is there, it is up to the profit interest to use it. And before regulations come, before anybody can draw the line, it is there to be misused. With respect to LLMs and artificial general intelligence, once again, Laurent calls it very much the AI’s ability to predict the next word.

Laurent: And about what you mentioned about the company that uses AI to choose what to publish and what not to publish, it very much reminds me of Netflix where when you are watching a series it is exactly at the seventeenth second that there is something that puts tension into the narrative, because they have all the statistics of what people like and what people don’t like, and when they switch off the TV and so on, but they all look alike if you notice. I am afraid if we let the machine decide we may end up with series that are uniform and that may entertain us for a while but where the element of surprise will completely disappear. I don’t think you can use it to imagine something new, something original that can have some impact years later.

Can AI be used to combat climate change?

Laurent: It is a challenging question. AI itself needs more resources to build bigger and bigger models, to crank more and more data, it already needs some of the largest supercomputers in the world. Some of the supercomputers in the world are being used just for AI at the moment, and in a way, I think it is part of the problem. Of course, AI is being used to fight climate change, to optimize energy consumption, on how to build more efficient batteries, AI is revolutionising science. There are many discoveries that is AI driven. So I am quite positive that AI will be one of the tools that will be used to combat climate change, but climate change won’t be solved by AI itself.

Appupen: We know we can’t solve the problem, so we like to put it on God, and I think now AI has taken on that role. It is part of the narrative to advertise AI’s role in fighting climate change, and of course, it can be used to do it because AI is being used at so many levels, but we need to reach a certain momentum to do so, and we are not there yet. I think there are many more areas where AI will be useful, but a lot of the money and effort is being invested in only profit-oriented areas.

So what are those areas where you think AI can be useful?

Appupen: Solar panels, minerals, protein folding and genetics, we have touched upon some of these in the book.

For Appu, I have this question. How did you come to choose the style in which you drew the comic?

Appupen: I try to change the style of my art depending on the story. In this case, the focus was more on the writing. And while I was writing, I realised I had to be adopt a pretty straightforward style of drawing. If I drew it in a cartoony style, the reader might not have taken the message of my book very seriously. If I drew it in a dark, adventure style, again, they might have read this as a thriller and nothing more.

Who is your favourite character in the book?

Laurent: I think, it’s Aida, the digital assistant. It’s he AI character, yes. It can be a bit annoying at some times, and we make it more human than it should be, but it is a friendly reminder of what AI can be.

Appupen: For me it is Super Hugo. I can break out of reality with Super Hugo. Sometimes he is very bare-faced and honest. I enjoyed delving into fantasy with that character.

What is your message to youngsters with respect to AI?

Laurent: Learn not only to use the technology, but also how it is being created, what is behind it, and what the limits of the technology are. We sometimes think that the answers from ChatGPT cannot be wrong while sometimes they can be plain wrong, so have your critical mind active while studying with AI.

Appupen: We are currently selling AI saying that it is going to make your life more convenient, but for whom and at what cost? The AI is trained to please you. This, I see, as an extension of the social media world. We see only what we like. We have a like but no hate button, so we are coaxed into a bubble where only what we want is coming into our feed. So slow down, the information you are getting may be twisted. Go deeper.

Since this book is set in a dystopia, how soon until the end of the world?

Laurent: I wouldn’t say myself the book is about an apocalypse. It has a fantasy ending where we ask the AI to imagine an apocalyptic setting. But we don’t need AI to take us to an apocalypse, we can do it very well ourselves.

Appupen: I have grown up reading and enjoying many stories about the apocalypse, be it being overtaken by new monsters or aliens or climate change, but I prefer the more dystopian ones where it is like a corporation or a country taking over; it seems more plausible for me that we go in that direction. Although reading such a story doesn’t mean it is going to happen, when we assessed the threat of AI we found that it is very possible for things to go in such a direction and we had to just shine a light on that view. It is a warning, it is what we can do as writers and artists.

But in the next ten years that is not going to happen, right?

Laurent: I think we have to be very humble when it comes to predicting the future of AI. It is going at an extreme speed at the moment and if I had told you earlier about the Hollywood writers’ strike that did happen, you would have said that I am completely crazy. Though always, always, it is not necessarily the AI that is taking your job, it is the person behind AI who is taking it.

Appupen: Well, your apocalypse could be simply economic for the innocent. Your end-of-the-world situation could simply be that you run out of resources! We use all the resources to train AI, and the AI takes care of a tiny bunch of people, and the whole world is rendered useless.