Unsupervised Learning with Jacob Effron
Unsupervised Learning with Jacob Effron

Ep 90: AI Pioneer Jürgen Schmidhuber on the State of AI Today

July 9, 2026

AI Summary

5 min read

“If you wait five years, compute is going to be ten times cheaper. You will be able to do the same thing for one tenth of the price.” That is Jürgen Schmidhuber’s blunt assessment of the trillion-dollar AI capex boom. Speaking on Unsupervised Learning, the AI pioneer—often called the father of modern deep learning—argues that the current investment frenzy in data centers and GPUs is a massive misallocation of capital. But his critique runs deeper than just business models. Schmidhuber, who has been working toward building an AI smarter than himself since the 1970s, believes the field is overlooking a fundamental problem: true artificial general intelligence (AGI) requires physical hardware that can match the human body, and we are nowhere close.

The Physical AI Bottleneck

Schmidhuber draws a sharp distinction between AI behind a screen and AI in the real world. Large language models, he notes, have now passed the Turing test, but that is not the same as AGI. “You can have a superhuman chess player behind the screen and something that passes the Turing test. But if it does not master the real world, it’s not an AGI. It’s just a fancy text editor.”

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

What you'll learn

  • 1 (01:34) **Defining True AI: More Than a Chatbot** - Schmidhuber argues that passing the Turing test is not enough; real AI requires physical embodiment and mastery of the real world.
  • 2 (03:40) **The Path to Recursive Self-Improvement (RSI)** - Schmidhuber traces his decades-long work on self-improving systems, from meta-evolution in 1987 to the mathematically optimal "Gödel machine" in 2003.
  • 3 (09:21) **On the Gradual vs. Sudden Nature of AI Takeoff** - Schmidhuber argues that from a cosmic perspective, the emergence of AI will look instantaneous, but the lived experience will be a gradual, continuous process of automation.
  • 4 (12:55) **The Key Limitation of Today's Large Language Models** - Schmidhuber criticizes the heavy reliance on pre-training on human-generated data, arguing it creates a super-human bias and ignores the vast majority of possible data.
  • 5 (16:30) **The Future is Artificial Scientists Driven by Curiosity** - Schmidhuber outlines his vision for AI that invents its own experiments and learns from the data they generate, driven by a formal theory of fun and creativity.
  • 6 (22:07) **Current State of AI Scientists and Chemistry** - Schmidhuber states that simple AI scientists already exist for specific applications, like chemistry, but haven't had their "ChatGPT moment" yet.
  • 7 (25:15) **The Hardware Bottleneck for Physical AGI** - Schmidhuber argues that the primary blocker for useful home robots is not just software, but the immense complexity and capability of the human body, which has no man-made equivalent.

+ Full timestamped outline available in the app

Show Notes

Dr. Jürgen Schmidhuber, a renowned scientist and AI researcher widely regarded as one of the pioneers in the field, originated key ideas behind today's transformers, LSTMs, and recursive self-improvement through his lab's work. He argues that true AGI remains bottlenecked by physical hardware, that today's AI data center investments are headed for a correction as open-source keeps pace with closed labs, and that the path to general intelligence runs through artificial curiosity and self-generated experimentation rather than internet data. He closes by reconsidering mainstream AI safety arguments and offers a sweeping vision of self-replicating robot societies eventually colonizing the solar system.

 

(0:00) Intro

(1:24) How Close Is Superhuman AI?

(2:27) Why ChatGPT Didn't Surprise Him

(3:21) The Path to Recursive Self-Improvement

(9:01) Will AI Takeoff Feel Sudden?

(11:02) Intelligence Means Efficiency

(12:32) Advice for Labs: Beyond Human-Biased Data

(17:10) Artificial Curiosity and the Theory of Fun

(21:33) When Do We Get the AI Scientist?

(24:07) AI Chemistry, MOFs, and Carbon Capture

(25:04) Robotics Reality Check

(28:23) The Data Center Bet: Overbuilt?

(31:48) Open Source vs. Closed Labs

(34:25) Does Being First to RSI Create a Moat?

(38:06) AI Safety and Alignment Skepticism

(43:44) Quickfire

 

With your host:  

@jacobeffron  

- Managing Director at Redpoint

Unsupervised Learning with Jacob Effron

More from this podcast

Unsupervised Learning with Jacob Effron →