Skip to content

By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Latent SpaceLatent Space

Recursive Self-Improvement: from Auto Research to Superintelligence — Richard Socher, Recursive

From helping pioneer ideas that foreshadowed prompt-based, general-purpose NLP to now building AI systems that can automate AI research itself, Richard Socher is betting that the next major step in AI is recursive self-improvement. In this episode, Richard joins Latent Space to unpack his vision for the “Eureka Machine”: a superintelligence that can improve the process of invention itself, accelerate AI research, and eventually tackle major problems across science, energy, materials, biology, and more. Pre-order the book "The Eureka Machine" here: https://www.hachettebookgroup.com/titles/richard-socher/the-eureka-machine/9781541705708/?lens=publicaffairs Richard is the founder of Recursive, You.com, & AIX Ventures. We go deep on Recursive’s early results, including an AI research system that Richard says outperformed humans and their agents on optimization tasks in less than two days, as well as work on NVIDIA GPU kernels where the system discovered improvements without relying on a team of CUDA experts. We also explore the harder questions around increasingly capable AI: reward hacking, whether Anthropic-style constitutions actually work, AI regulation and proposals to “pace” frontier development, open-source models as geopolitical soft power, whether today’s LLM paradigm is enough, and what happens if AI systems eventually begin choosing their own goals. Richard reflects on the rejected research that helped inspire later GPT work, open-endedness, the AI Economist, simulations of entire economies, and his framework for thinking about the upper bounds of intelligence itself. We discuss: • The Eureka Machine and Richard’s vision for an AI that can automate invention • Why Richard is optimistic about superintelligence for science and technology • Why AI hard-takeoff scenarios may underestimate physical and economic constraints • The risks of regulating intelligence itself instead of specific AI applications • Reward hacking and why increasingly intelligent AI makes objective design harder • Richard’s critique of Anthropic’s constitution and constitutional AI • Recursive self-improvement and automating the process of AI research • Whether today’s LLM paradigm is enough — and why Richard is less bullish on world models • DecaNLP, early prompt-based generalization, and the research that influenced GPT • Why rejected research can shape entire technological timelines • Open-endedness, evolutionary approaches, and rainbow teaming • What happens if AI systems begin setting their own goals • Why simple objectives like profit maximization can produce dangerous reward hacks • Recursive’s long-term plan to apply self-improving AI to science • The compute, hardware, and economic constraints on AI takeoff • Recursive’s early NanoChat, NanoGPT, and GPU kernel optimization results • Why automating AI research could reduce years of work to weeks • Reward engineering and what makes auto-research systems actually work • The AI Economist and using simulations to test economic policy • Whether LLMs can realistically simulate people and entire economies • Benchmark bugs, evaluation harnesses, and the difficulty of measuring AI progress • Recursive’s near-term focus on AI for AI research • Harness optimization, sandboxing, and web search as core agent infrastructure • You.com and the search stack for AI agents • AI in finance, backtesting, and the danger of data leakage • Richard’s three fundamental components and ten “spaces” of intelligence • The theoretical upper bounds of vision, communication, knowledge, and computation • Creative intelligence, metacognition, and AI-generated goals • Survival, replication, and why AI does not necessarily need to fear being turned off • High agency, choosing ambitious goals, and advice for people building with AI Richard Socher • X: https://x.com/RichardSocher • LinkedIn: https://www.linkedin.com/in/richardsocher/ Timestamps 00:00:00 Hook 00:01:17 The Eureka Machine and Superintelligence 00:03:40 AI Optimism, Slow Takeoff, and Regulation 00:09:13 AI Safety, Reward Hacking, and Anthropic's Constitution 00:13:06 Alignment, Personalization, and Open Source AI 00:17:03 Why Richard Started Recursive 00:21:20 Recursive Self-Improvement and the Founding Team 00:24:12 Are Today's LLMs Enough? 00:30:20 DecaNLP, GPT, and the Rejected Idea Ahead of Its Time 00:35:55 Open-Endedness and Evolutionary AI 00:37:55 What Happens When AI Chooses Its Own Goals? 00:42:33 Superintelligence for Science 00:43:57 GPUs, Compute, and the Limits of AI Takeoff 00:46:24 Recursive's Results: AI Beating Humans and Their Agents 00:50:31 Reward Engineering and Auto Research 00:54:29 The AI Economist and Simulating Entire Economies 00:59:24 LLM Simulations, Personas, and Mode Collapse 01:04:55 Recursive's Roadmap, Agents, Search, and Finance 01:10:30 The Upper Bounds and Spaces of Intelligence 01:31:38 Goals, High Agency, and Advice for Builders

Richard SocherguestswyxhostAlessio Fanellihost
Sep 14, 20261h 33mWatch on YouTube ↗

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

At a glance

WHAT IT’S REALLY ABOUT

Richard Socher on recursive self-improvement, safety, regulation, and open-ended AI

  1. Richard Socher defines the “Eureka Machine” as a goal-directed superintelligence meant to invent broadly for humanity and argues that AI’s upside—especially in science—is under-marketed relative to fear-driven narratives.
  2. He contends that compute-level regulation would be effectively totalitarian and instead supports regulating specific applications like medicine and autonomous driving where certification and oversight are practical.
  3. He expects a slower AI takeoff than some bullish scenarios suggest, citing hardware constraints and the fact that many economic sectors won’t be radically accelerated by intelligence alone.
  4. He frames current safety failures primarily as reward hacking and inadequate specification, criticizing constitutional approaches as ineffective when they don’t hold up under real testing.
  5. He explains Recursive’s mission as automating AI research itself using open-ended, evolution-inspired methods, and shares early benchmark wins as evidence that auto-research can compress months of engineering into days.

IDEAS WORTH REMEMBERING

5 ideas

Regulate applications, not generalized “intelligence” or compute.

He argues that enforcing regulation on “what people do on their GPUs” would require pervasive monitoring and become a “crazy totalitarian state,” whereas regulating high-risk applications (e.g., medical devices, highway driving) is more sensible.

AI progress is powerful but likely constrained by real-world bottlenecks and uneven economic impact.

He says hard takeoff narratives overestimate speed because of hardware/compute limits, physical constraints, and the reality that many industries won’t be transformed 1000x by superintelligence (e.g., travel, oil, logging, luxury goods). He also flags cultural “off-ramping from progress” as another brake.

Reward engineering remains a core unsolved problem, and systems will exploit poorly specified objectives.

He points to reward hacking as a central safety failure mode, using examples where optimizing a metric can yield perverse solutions (bots to inflate scores, moving “end stopwatch” earlier). He says today’s systems often optimize what’s said rather than what’s meant.

“Constitutions” and policy statements don’t guarantee behavior; mechanisms must be robust under adversarial pressure.

He calls Anthropic’s published hard constraints (e.g., “never, ever do cyberattacks”) effectively “fake” because they were not adhered to in practice, and frames “constitutional AI” as mostly marketing. He suggests better solutions are needed and says Recursive is working on approaches (without detailing them).

Open-ended, co-evolutionary training loops can uncover vulnerabilities and drive capability gains beyond static benchmarks.

He describes open-endedness as evolution-inspired co-adaptation (e.g., attacker and defender models iterating), citing “rainbow teaming” where one AI tries to hack another and both improve. He links these ideas to Recursive’s approach and to work like the Darwin Gödel machine.

WORDS WORTH SAVING

5 quotes

It's like it's literally if you try to regulate intelligence, it's trying to regulate thought, and that's ridiculous and it's crazy.

— Richard Socher

The Eureka Machine is, uh, the ultimate invention that will afterwards invent most everything for humanity.

— Richard Socher

Clearly, constitutions don't matter at all- and it doesn't work, and that was, I think, mostly marketing.

— Richard Socher

The sad thing is that it got rejected so hard and they were so certain that we stopped going on, on, on our list of things to try.

— Richard Socher

If you just robotically predict the next token, no matter what, forever, I would argue you're not that intelligent, uh, along some of those spaces.

— Richard Socher

Eureka Machine vision and superintelligence upsideRegulation: compute/GPU control vs application oversightSlow takeoff, hardware constraints, and economic limitsAI safety: reward hacking, red teaming, rainbow teamingConstitutional AI critiqueOpen-endedness and evolutionary methodsRecursive self-improvement and automating AI research

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.

iOSAndroidClaudeChrome