Bloomberg's Joe Weisenthal and Tracy Alloway analyze the weird patterns, the complex issues and the newest market crazes. Join the conversation every Monday and Thursday for interviews with the most interesting minds in finance, economics and markets.

Odd Lots

In healthcare, finance, and even scientific research, the most advanced AI models are useless if regulators and the public don't trust them. Building that trust — through interpretability, testing, and oversight — may be a harder problem than model accuracy.

The biggest barrier to AI adoption is trust, not technology.

Beyond the headline-grabbing protein structure predictions, the deeper impact of AlphaFold is the way it turns biology into a more computational, iterative science — and that shift has huge implications for the biotech economy.

AlphaFold's real legacy is transforming how biology research gets done.

The immense energy and compute demands of frontier AI training mean that AI supercomputers are now strategic assets that shape energy grids, supply chains, and industrial policy. This isn't just a tech story — it's a macro story.

AI infrastructure is becoming critical national infrastructure.

As compute costs and energy demands explode, what happens when the traditional scaling laws that have powered AI progress start to plateau? How should the industry think about the economics of the next generation of models?

Is the AI scaling era running into physical limits?

From AlphaGo to AlphaZero, DeepMind has used reinforcement learning to conquer games. What other real-world problems are structured enough for RL to deliver similar breakthrough results, especially in chemistry or materials science?

Can reinforcement learning unlock discoveries beyond games?

As AI systems get more general and more autonomous, what does responsible governance look like? How do we balance open research with the dangers of misuse, and what role should companies like DeepMind play in setting those boundaries?

How should democracies grapple with the national security risks of AGI?

Demis Hassabis has often said the goal is to build AI that can generate its own ideas and hypotheses. What would it take to convince a skeptical scientist that an AI's output is genuinely novel thinking, not just pattern completion?

What empirical test would convince you an AI is truly creative?

Dive into the cutting-edge worlds of technology and science with Dwarkesh Patel's meticulously researched interviews that bring together brilliant minds to unpack revolutionary ideas in AI, biotech, and beyond. Each episode delivers profound, thought-provoking conversations that challenge conventional wisdom and illuminate the paths shaping humanity's future.

Dwarkesh Podcast

A Neuroscience podcast. Each week I talk with experts on a range of topics from consciousness and neuroplasticity to evolution and everything in between. I have a particular interest in consciousness and understanding our inner subjective conscious experience. Evan has a Neuroscience degree from Trinity College Dublin and is the co-founder of 2 (failed) edtech startups. He is a professional tennis coach and Triathlete enthusiast.

Get Demis Hassabis on Odd Lots

The Case for This Conversation

Shape What Gets Discussed

You Might Also Enjoy