Key Takeaways from the World AI Cannes Festival

The French city of Cannes, known more for its renowned film festival than an AI conference, nevertheless played host this week to around 16,000 AI experts from around the world.

For three days, this French Riviera city was taken over by the annual ‘festival de la technologie’ that is the World AI Cannes Festival (WAICF).

More than a year after the hype of ChatGPT, the 2024 conference promised to be a more down-to-earth and mature look at AI.

AI Business breaks down the biggest themes and talking points from the show.

A real maturing

AI at this event was like what the French call ‘entre deux’ in that it has matured compared to previous iterations.

Speakers were not spouting things like, ‘look what you can do with ChatGPT.’ Instead, speeches and panels had more nuance. Discussions were often high-level, with regulation, ethics and implementation strategies focusing on real tangible applications.

For example, Jean-Bernard Hentz of Airbus said AI can improve efficiency, quality and safety in manufacturing. And Anna Mareschi Danieli from steel company Danieli & C S.p.A said AI tools can allow workers to use their brain for “more important thinking, for value-added activities.”

Conversations focused on finding ways and ideas to improve AI, with no one really happy with how the technology is at present. Speakers wanted more focus on developing safer, energy efficient systems – a level of maturity and realization that often is shoved to the wayside in favor of showing levels of performance.

Related:Meta’s Yann LeCun Wants to Ditch Generative AI

Yann LeCun on display

Meta’s Chief AI Scientist Yann LeCun was in fine form this year and continued his annual challenge of the norms of the current state of the AI industry. This year, he attacked the current state of machine learning, auto-regressive models, and the EU AI, act among many other things. 

He commanded a full house across two stages with attendees hoping to catch a glimpse of one of AI’s godfathers.

Meta's Yann LeCun

A particular highlight was the combative panel discussion where he appeared on stage with Mark Brakel from the Future of Life Institute – the group that called for a six-month pause in advanced AI development − which made for memorable encounters and arguments about human cloning, nuclear weapons  and the dangers of kitchen knives, of all things.

From claims of wanting to generate AI and that the current state of machine learning “sucks,” the 2024 edition of LeCun in Cannes did not disappoint.

The only absence, however, was the baying mob of developers, hoping to talk to the Turing award winner, which was discouraged this year. The stage chair said LeCun’s 2023 keynote was the closest he had ever seen to a post-speech riot at an AI conference.

Related:NeuroAI: Tapping the Human Brain to Fuel AI Development

There’s always next year though.

A new way of thinking about AI

One of the major themes that arose in Cannes was doing AI differently. Speakers from across the two days spoke of the need to evolve AI to make it low-cost and improve its abilities.

One concept that arose was the need to move on from just language. Yann LeCun, Dileep George from Google DeepMind and Professor Giacomo Indiveri all gave talks on the potential of enabling AI to mimic how our brains learn to improve its training.

George, DeepMind’s director of research, said that even if humans did not have language, they would learn from their experiences and apply that knowledge to their actions.

Google DeepMind's Dileep George

The various talks on learning from the brain - or NeuroAI - all agreed that current hardware limitations prevent AI from powering use cases seen in science fiction, like Jetson’s Rosie the Robot.

Tristan Stöber from Ruhr-Universität Bochum spoke of how neuromorphic hardware could allow for a neural network to be directly ingrained onto a piece of hardware, a concept that could improve efficiency far greater than any system we use today.

Related:AI vs. Nuclear Weapons: Debating the Right Analogy for AI Risks

It is a long way to go to achieve anything close to the AI brain concepts proposed during the Cannes conference, but having experts wanting to push beyond the endless scaling of systems was a welcome glimpse into the potential future of AI.

Open source, open mind

One of the consensus among speakers was that AI needs to be open source.

Luc Julia, the co-creator of Siri and Renault’s chief scientific officer, said that while open source was not the answer to everything, it made AI more transparent.

“There are more chances that you can verify what is the data that was used, what are the algorithms that are being used, and it's better way to have more control,” he said.

Renault's Luc Julia

Hugging Face’s chief evangelist, Julien Simon, said that open source keeps closed model developers on their toes. And AMD's Ramine Roane said the only way to make progress in AI is through “open teamwork.”

The fact that open source was a major talking point across panels showed that attendees and speakers did not want the handful of tech companies that currently dominate the AI landscape to continue to do so.

In the worlds of LeCun, the future of AI “must be open source.”

Elsewhere at WAICF

Most AI conferences have a defining concept that brings together the organizations there, such as a specific sector or profession like accounting. But the WAICF continued its streak of bringing together a truly wide array of firms.

This year’s eclectic gathering included generative AI startups, a host of hardware firms and the big-name players one would expect at an AI conference − AMD, Dell and HPE were all present.

But even Tesla surprisingly had a spot – a very understated spot, hidden away under a flight of stairs with just a flag, a car and an Optimus Bot to showcase that they were even there.

Optimus bot

The show floor was littered with robots. The likes of Buddy, the emotional robot from Blue Frog and United Robotics Group’s server robot, Plato, roamed the isles causing passing attendees to stop and gawk – causing the odd irritating pile-up.

Given the event opens to the general public on the weekend, gadgets like these are more likely to wow the average Joe in the street, but even seasoned AI experts were impressed with the interactive demos.

Even on display was Apple’s brand new Vision Pro headset. Staff from Milan-based startup dilium were putting the mixed reality device through its paces, showing how users can interact with everyday objects and tasks, like reading the paper, while wearing the device.

That is what makes Cannes so enjoyable as an event, it is a hodgepodge of various AI groups coming together against the backdrop of the French Rivera. While it has no defining individual vision as such, the World AI Cannes Festival combines high-level conversations with shiny gadgets, so it has something for everyone. The view from the conference center was not half bad either.

It is Cannes, after all.