Les podcasts Sparks du CERN explorent l'intelligence artificielle

View in


Les podcasts Sparks du CERN explorent l'intelligence artificielle

lun 19/07/2021 - 10:08


CERN is launching a new podcast series on artificial intelligence. The series looks forward to the first edition of the Sparks! Serendipity Forum in September, when over 30 leading thinkers will converge on the laboratory for high-level multidisciplinary discussions designed to spark ethical innovation.

To whet your appetite for the forum, the podcasts bring a selection of the Sparks delegates together in pairs. Think of these conversations like collisions in the LHC. Rather than protons, however, we’re colliding a diverse cast of thought leaders featuring computer scientists, neuroscientists, philosophers, physicists and psychologists. And rather than energy being transformed into new particles, the weight of scholarship is transformed into new seams of creative energy — a foretaste of the intellectual dust-up planned for September.

The first episode, Brainy AI, wastes no time in colliding two giants of the field: renowned machine-learning expert Stuart Russell, and Tomaso Poggio, one of the founders of computational neuroscience. Our guests contrast the human brain with deep learning, and pull no punches in their evaluation of the direction of AI research. “What if we succeed?” asks Russell. “AI has been so difficult for so long, that we kind of forget to ask that question.”

AI for science
Quantum AI keeps up the momentum by colliding neuroscientist and AI entrepreneur Vivienne Ming with quantum-computing expert Maria Spiropulu. The pair connect the two ongoing revolutions in computing, and explore links between neuroscience and quantum physics. “To see that our brains and the universe are solving the same problems mathematically in the same ways, I think is really exciting,” says Ming, “ — particularly if you’re a little demented like me, and want to build cyborgs and give vision back to the blind.”

The two central pillars of our series of six half-hour podcasts relate directly to the fundamental research we do here at CERN. First up, world-renowned theoretical physicist John Ellis talks to leading AI researcher Anima Anandkuma about how to design and deploy Creative AI. Then, in Experimental AI, we invite you to listen in on a conflab between two of CERN’s finest: Michael Doser, of the antimatter factory, and Maurizio Pierini, of the CMS experiment. Whether in the exquisitely creative process of theory formation, or the nuts-and-bolts work of making reliable experimental measurements, our guests agree that artificial intelligence is an irresistibly powerful tool for particle physics, both now and in the future.

We arrive at the crux of the issue for many in Episode 5: how can we design Ethical AI? Philosopher S. Matthew Liao and computer scientist Nyalleng Moorosi bring a laser focus to their analysis of ethical challenges posed by real-world AI, from autonomous weapons to everyday algorithms used by mid-sized companies. Whatever the subject, the decisions made by deep learning are only as good as the data they work with, and risk being deficient in fairness and accountability. “When we had these decisions being made by humans, who we could embed a whole history book inside, perhaps there would be areas of more flexibility,” says Moorosi. “But the moment we start to encode these things, we are starting to encode some of these misfortunes, and we’re going to start to perpetuate them.”

Is there anything that humans can do that AI cannot in principle do?

Daniel Kahneman

Last but not least, we collide AAAI President Elect Francesca Rossi with Nobel laureate Daniel Kahneman — one of the greatest living cognitive psychologists. In Fast and slow AI, our guests take Kahneman’s revolutionary systems of thought as inspiration for rewriting AI, and debate the nature of thought itself. “I really find it difficult to imagine why there should be anything at which humans are essential in the domain of intelligence,” says Kahneman. “Is there anything that humans can do that AI cannot in principle do?”

Artificial intelligence is transforming our world, and nobody knows for sure what the future will hold. Hosts Abha Eli Phoboo and I invite you to join us on a journey of discovery as we invite the great and the good of this fast-evolving field to share their wisdom in a spirit of open collaboration, ethically minded innovation and indefatigable curiosity — the finest traditions of CERN.

Sparks is a CERN outreach event and part of the CERN & Society programme. CERN & Society activities are only possible thanks to voluntary support received from partners, in particular Rolex and its long-standing association with the Organisation. The 2020-2021 Sparks is also supported by Edmond de Rothschild and by the Didier et Martine Primat Foundation.