Video · 2026-07-10 · 1h 4m · 6 moments

Open Source Wins, AGI Is Here, and Scorsese’s AI Toolkit with CEOs of Cerebras & Black Forest Labs

✦ AI generated

GUEST Robin Rombach GUEST Andrew Feldman GUEST Jason Calacanis

timeline · colored by role

Data◆

The current AI data center buildout will consume more power in the next several years than the world used in the previous 50 years, with individual buildings drawing more electricity than midsize cities.

Feldman describes the scale of the ongoing global AI data center buildout as unprecedented, comparing its power consumption to more than the last 50 years combined.

transcript

Andrew Feldman: What we're talking about now are data centers that are in the next several years going to use more power than the previous 50 years on Earth took. We're talking about individual buildings the size of football fields that have more power coming into them than midsize cities.

provides context · 1

Claim

The US needs more domestic open-source AI models so companies aren't forced to choose only between OSS 12B and Chinese models.

Feldman argues that while OpenAI's open-weight release was a good step, America needs more homegrown open-source models so regulated industries and sovereignty-minded companies aren't limited to OSS 12B or Chinese alternatives.

transcript

Andrew Feldman: I think OpenAI made a good call releasing OSS 12B some months back. That was a good open-source model. But I think in the US we need more domestic open source models. We need to give the world a choice. If they want to run open source right now, it's OSS 12B or Chinese models.

extends · 1provides context · 2supports · 6

Claim◆

By any definition used 20 or more years ago, we have already achieved artificial general intelligence, even though it hasn't been fully deployed yet.

Jason argues, with Andrew's agreement, that current AI systems already meet any definition of AGI used decades ago, including the Turing test, even though it hasn't been fully rolled out into the world.

transcript

Jason Calacanis: And AGI I think I suspect you'll agree with me that we've hit it. We just haven't exactly deployed it fully. We have artificial general intelligence now... by any definition we had 20 years ago we've hit it.

rebuts · 2supports · 1

Mechanism

Recursive self-improvement loops in AI, where a model learns from its own results and repeats, produce vastly better answers each cycle rather than marginal gains, and it's unclear where the exponential curve stops.

Feldman explains that the recursive gains Sam Altman, Ilya Sutskever, and Dario Amodei foresaw years ago, where a model learns from its own output and iterates, compound exponentially rather than linearly.

transcript

Andrew Feldman: Powerful recursive gains are exponential — you get better, you do it again. And if you continue to get gain, the slope of that curve is so steep. You ask it a question, you learn from the results, you ask it to do it again, the results get better and more information is added. These sort of loops are producing not a little bit better answers but vastly better answers.

extends · 1

Mechanism

The same generative model that makes images and videos can be extended with action prediction so it can ultimately be deployed as the controlling brain on a physical robot.

Rombach describes Black Forest Labs' shift toward multimodal models that combine image, video, and audio generation with action prediction, enabling deployment on real-world robots.

transcript

Robin Rombach: We are now like entering a new paradigm which is combining that with something that's called action prediction such that you can actually use the same model to make images, to make videos, to make audio, and to predict actions — which means you can ultimately deploy it on a robot in the real world.

Claim◆

AI generative models are a medium, not a prescribed workflow — creators like Martin Scorsese should be free to use them however they envision, without being told how.

Rombach describes his philosophy behind partnering with Martin Scorsese: AI models are a medium meant to be shaped by the artist's vision, not a tool that dictates how films should be made.

transcript

Robin Rombach: These AI models, they are a medium, right? We don't want to set any way of how they are supposed to be used. We don't want to tell anyone — especially not someone like Martin Scorsese — how he is supposed to use these models. He is one of the greatest filmmakers ever.

Highlight slides

AI Data Centers: An Unprecedented Power Surge✦ from: The current AI data center buildout will consume more power in the next several years than the world used in the previous 50 years, with individual buildings drawing more electricity than midsize cities.Buildings Bigger Than Cities' Power Needs✦ from: The current AI data center buildout will consume more power in the next several years than the world used in the previous 50 years, with individual buildings drawing more electricity than midsize cities.We've Already Hit AGI✦ from: By any definition used 20 or more years ago, we have already achieved artificial general intelligence, even though it hasn't been fully deployed yet.Achieved, Not Yet Deployed✦ from: By any definition used 20 or more years ago, we have already achieved artificial general intelligence, even though it hasn't been fully deployed yet.AI Models Are a Medium, Not a Method✦ from: AI generative models are a medium, not a prescribed workflow — creators like Martin Scorsese should be free to use them however they envision, without being told how.Trusting the Artist's Vision✦ from: AI generative models are a medium, not a prescribed workflow — creators like Martin Scorsese should be free to use them however they envision, without being told how.

Related episodes

Video · The Cognitive Revolution

Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation Models

→

Video · The Cognitive Revolution

LTX CEO on Video Gen + Can we beat AI superforecasters?

→

Video · 20VC

Should the US Ban Chinese Open-Source Models | OpenRouter's Chance To Sell | Stripe Buying PayPal

→