Artificial Intelligence and You

This and all episodes at: https://aiandyou.net/ .

We're going underwater, to talk with whales. There have been some amazing breakthroughs recently in understanding the language of sperm whales, and AI plays a fascinating part in that. And even if you have no immediate intentions to go say hi to a whale, AI's part in this story has ramifications far outside the field of cetacean communication. Taking us through that is Gašper Beguš, whose recent research has produced one of the most startling advances yet in the quest to understand whale communication.

Gašper is Associate Professor of Linguistics at UC Berkeley, where he focuses on interpretable AI, combining linguistics, cognitive science, machine learning, neuroscience, philology, and marine biology. He is the Linguistics Lead at Project CETI, a nonprofit organization applying advanced machine learning and state-of-the-art robotics to understanding whales, working in Dominica in the eastern Caribbean. In his Berkeley Biological and Artificial Language lab, he also builds the most realistic models of human language learning. He's been featured everywhere from the BBC and the Guardian to Time, The Atlantic, and National Geographic.

We start with a fascinating backgrounder in linguistics, then talk about what's captivating about whales, comparing them to other vocalizing creatures, talk about communicating with aliens and interspecies communication, what language even means for whales, and… the birds and the bees. 

All this plus our usual look at today's AI headlines!

Transcript and URLs referenced at HumanCusp Blog.

This and all episodes at: https://aiandyou.net/ .

I am talking with Katja Grace, founder of AI Impacts, a forecasting and risk research project focused on AI safety and governance. She has leveraged her passion and concern about AI risk into a significant role in, for instance, surveying the AI community for its evaluations of risk, which got her named to Time Magazine’s 100 most influential people in AI for 2024. She is an expert on technological trajectories, model scaling, and anthropic reasoning.

In our conclusion, we talk about predicting superintelligence, Katja’s p(doom), regulation, the Great Filter argument, and Katja’s childhood calculations for dealing with lethal snake encounters.

All this plus our usual look at today's AI headlines!

Transcript and URLs referenced at HumanCusp Blog.

This and all episodes at: https://aiandyou.net/ .

When you stop and really consider the existential risks of AI, it becomes very apparent that we should do something. Yet so few people do. Someone who’s attempting to make up for that is Katja Grace, founder of AI Impacts, a forecasting and risk research project focused on AI safety and governance. Her surveys of the AI community’s evaluations of risk have gotten her named to Time Magazine’s 100 most influential people in AI for 2024 and her papers on those surveys have been some of the most discussed scientific papers of their time. She’s an expert on technological trajectories, model scaling, and anthropic reasoning.

We talk about what’s driving Katja to do this work, the fundamental risks that she sees in AI, AI evolving unexpected goals, risks of agents, comparing unemployment risks and like extinction risks, artificial general intelligence, the stunning results of the surveys she’s run on when AI will have serious effects, and… Trojan Horse racing.

All this plus our usual look at today's AI headlines!

Transcript and URLs referenced at HumanCusp Blog.

This and all episodes at: https://aiandyou.net/ .

I have been talking with Tomaso Poggio, Eugene McDermott professor in the Department of Brain and Cognitive Sciences at MIT and the Director of the Center for Brains, Minds, and Machines, and one of the founders of the field of computational neuroscience. Tomaso is a fellow of the American Academy of Arts and Sciences and of the American Association for the Advancement of Science, and a founding fellow of the Association for the Advancement of Artificial Intelligence. He develops models of brain function that illuminate human intelligence and builds intelligent machines that can mimic human performance His new book, Brains, Minds, Machines, The Mystery of Human Intelligence, the Enigmas of the Artificial, comes out this summer.

We talk about learning in the brain and synaptic mechanisms, the role of sleep, what AI scientists should pay more attention to from neuroscience, other computational mechanisms in the brain besides neurons, connectomics, robotics, and… flies and worms.

All this plus our usual look at today's AI headlines!

Transcript and URLs referenced at HumanCusp Blog.

This and all episodes at: https://aiandyou.net/ .

Studying human intelligence is a matter of neuroscience, and creating software is a matter of computing, so creating artificial intelligence would be at the intersection of those fields, called computational neuroscience, and I have with me one of the founders of that field. Tomaso Poggio is the Eugene McDermott professor in the Department of Brain and Cognitive Sciences at MIT and the Director of the Center for Brains, Minds, and Machines. He is a fellow of the American Academy of Arts and Sciences and of the American Association for the Advancement of Science, and a founding fellow of the Association for the Advancement of Artificial Intelligence. His home page says that he “develops models of brain function that illuminate human intelligence and builds intelligent machines that can mimic human performance.” Wow. His new book, Brains, Minds, Machines, The Mystery of Human Intelligence, the Enigmas of the Artificial, comes out this summer.

Tomaso defines computational neuroscience, and then we talk about computation in the human brain, how large language models landed for him, holography, limitations of LLMs, and backpropagation equivalents in the human brain.

All this plus our usual look at today's AI headlines!

Transcript and URLs referenced at HumanCusp Blog.