Opportunities, limits and ethical issues: what scenarios for AI in 2025?

ChatGPT, Gemini, xAI, Claude – the list of conversational agents powered by artificial intelligence (AI) making headlines grows longer every week. Competing in a market projected to reach USD 1.34 trillion within the next five years¹, tech industry giants are jostling to develop the most advanced generative AI model. The stakes could hardly be higher: from defense sector contracts² and technological sovereignty to the transformation of economies and labour markets.

The pervasive nature of AI is both fascinating and unsettling. Where do we stand in this revolution? Can its development be controlled? Has AI reached maturity, and what are its next major milestones? To explore these questions, we spoke with Alexandre Pouget, Professor of Computational Neuroscience at the University of Geneva and a leading AI expert. His work spans science, foresight, and profound reflections on the role of artificial intelligence in human evolution.

How do you define AI?

Artificial intelligence is an attempt to reproduce all the cognitive capabilities of a human being on computer. In the early days of AI, the idea was to match human capabilities. Now people realise that they will be surpassed fairly quickly, at least in theory. The question is to know how to get there, for instance by reproducing the way the brain works and the way neurones talk to each other.

There has been a monumental leap forward in the past ten years… But this is just an intermediate stage on the way to a more general AI

In the 1970s there were attempts to reproduce our ability to reason by studying human beings, but without looking specifically at neurones. For 20 years now, there have been attempts to reproduce the capabilities of human intelligence, but this has been done by using the neural mechanisms of the brain functions.

At this time, what is your assessment of the current state of AI?

There has been a monumental leap forward in the past ten years. We are at or even beyond the human level for many tasks, like face and object recognition. But this is just an intermediate stage on the way to a more general AI that will be increasingly human-like and capable of greater autonomy. Time scales are much shorter these days, somewhere between 10 to 50 years; not that long ago people were talking in terms of many decades.

In which areas of everyday life do you anticipate the greatest impact over the next few years?

We are going to reach a stage where there will always be several helpers available to us. We will all have personal assistants to help us plan every aspect of our lives.

Anyone trying out ChatGPT can see that it is already a useful assistant. In the future, AI will become more personalised. It remains to be seen what sort of relationship we will develop with these assistants on a personal level.

But when it comes to robots, there are physical constraints. Robotics is not yet sufficiently far advanced. The other issue is that it uses a great deal of energy. The technology isn’t quite there yet in that respect. When it is, it will revolutionise entire industries, such as healthcare, with the use of assistants in nursing homes, for example.

It’s interesting that perhaps for the first time in its history the AI industry itself is asking to be regulated

The major economic blocs are currently battling it out over AI. Who’s making the most progress between the US, China and Europe? 

It is beyond dispute that the US is currently in a dominant position. The situation could become dangerous if China or Russia were to take the lead, given that these countries are governed by autocratic regimes.

In Europe, some people say we are being held back by regulation, but that’s not my impression. We’re certainly the first ones to create ethics charters for AI, but my colleagues and I don’t feel that these hold back our work in any way.

Read also: Rethinking through the noise | Lombard Odier

Turning to the US again, it’s interesting that perhaps for the first time in its history the AI industry itself is asking to be regulated. Even though OpenAI is doing the precise opposite of what it said it would do by becoming a for-profit organisation, you cannot imagine the number of publications that reveal the secrets of these systems every day, so I'm not too concerned.

What are the dangers AI might pose for human society, and what advantages might it bring?

The advantages are obvious. We are advancing one stage further in evolution. Artificial intelligence systems will far outstrip humans. Everyone always forgets that we are only a single moment in evolution; we are doing a little bit too much navel-gazing by assuming that humans represent the pinnacle of the evolutionary process.

Artificial intelligence systems will far outstrip humans. Everyone always forgets that we are only a single moment in evolution

AI systems will be extraordinary and capable of solving complex problems in physics or biology. Take a three-hour course on cancer. It’s incredibly complex. These are the things artificial intelligence will be able to model for us. In physics, AI will do colossal calculations on dark energy and dark matter.

Karim Lakhani, a professor specialising in workplace technology and AI at Harvard Business School, says, “AI won’t replace humans – but humans with AI will replace humans withoutAI.” Do you agree?

It all depends on what he means. If he's talking about AI assisting humans, it’s clear that if you don't know how to use AI, you’re going to be less competitive. But if he’s talking about hybrid humans with AI, it’s a possibility in the future, that’s for sure. But I'm not yet convinced that this will be the next stage.

From a biological standpoint, for example, the brain is a miracle of energy efficiency. That's the most interesting thing about it from the engineering perspective: how do you reduce energy consumption to the order of 100 Watts³ for a human body and 20 Watts for the brain?⁴ These are amazing figures in terms of energy efficiency for the system, and AI can’t match that today. Engineers would really love to reproduce this feat.

In the coming decades, the race will be on for personalised AI, and whoever has the best will win the lion’s share

AI is already widely used in the banking sector to combat fraud, sort data and detect anomalies, for example. Do you think it will replace bankers one day?

If you mean using AI as a tool for granting credit, for instance, why not? There’s also the competitive dimension, which is complex; if every institution is using AI, what's going to make one bank different from another? Initially, customer services in a branch will be transformed rapidly by AI.

our expertise.

rethink investments.

learn more

private banking.

learn more

For wealth management, AI will be assisting rather than replacing. But the question that has to be asked is this: what happens if everyone is using the same AI? Hedge funds, for example, will have to make the most of the skills the others don’t have, and the key thing will be to develop your own AI model. In the coming decades, the race will be on for personalised AI, and whoever has the best will win the lion’s share.

Do you see some sectors using AI better than others?

In medical research, or medicine in general, things are happening very quickly. If you are a doctor, for instance, start-ups are offering to record the conversation with your patient and the AI system will produce a summary of the consultation. The time saved is amazing.

Read also: Future-proof | Lombard Odier

In research, the writing is already done by AI, and so is the programming. Soon, it will be able to help us as we reflect on problems. Some areas already have specific applications, like AlphaFold, the artificial intelligence tool that allowed the winners of the 2024 Chemistry Nobel Prize to discover protein structures in record time.⁵

Outside your research, are there works that have inspired and influenced the way you approach artificial intelligence?

Very early in my career the book Neuronal Man by Jean-Pierre Changeux changed my life. In the 1980s, this was the first reductionist manifesto based on neuroscience that explains we are nothing more than a network of neurones, and that every aspect of our psyche flows from that. I was 19 years old at the time. Now, at the age of 57, I see when I talk about it today that most people are astounded and find it hard to accept the idea.

The film Blade Runner also asks all the big questions about AI and warns of the dangers it represents, both ethically and from the idea of humans being overtaken. Everyone needs to see this movie. It shows robots catching up with human capabilities, and what happens when this occurs.

There’s still an awful lot of uncertainty. Seven or eight years ago, very few people would have been able to predict the ChatGPT revolution and what it is able to do today

Imagine we are in 2035. What position will AI have in society? Will machines have enhanced us, or replaced us?

Looking ahead ten years, I find it hard to see machines replacing us, but AI will have transformed our daily lives. Assistants will be everywhere and available to help us on our smartphones. But I don’t think it will go any further than that.

There is already a degree of saturation today. ChatGPT and the like are starting to hit saturation because they have no “fresh” data to train on, and I haven’t yet seen the next big idea appearing. That would be autonomous systems designed with the ability to plan in order to achieve set objectives and organise themselves fully autonomously for that purpose.

You have to understand that the current revolution is based on ideas that are 30 years old, adapted onto more powerful computers. Is there a new idea, on a par with the ones that emerged in the 1980s, that would change the game? I haven't seen it yet. There's still an awful lot of uncertainty. Seven or eight years ago, very few people would have been able to predict the ChatGPT revolution and what it is able to do today.