Skip to main content
a16z Podcast

Can AI Learn to Improve Itself? | Recursive Self-Improvement

17 min episode · 2 min read
·
Mark Chen,John Schulman

Episode

17 min

Read time

2 min

Topics

Fundraising & VC, Artificial Intelligence, Software Development

AI-Generated Summary

Key Takeaways

  • ✓RSI Progress Spectrum: Mark Chen estimates AI is roughly 25–50% toward full RSI, defined as models autonomously discovering and implementing research ideas, compared to approximately 80% toward AGI. Anthropic data shows roughly a 2x research speed increase in one year, currently driven by human-AI collaboration rather than full automation.
  • ✓AI Research Intern Framework: Current models function as scoped research interns—capable of implementing, running, and reporting on predefined experiments end-to-end. They fail at open-ended ideation. Practitioners can deploy this today by assigning specific, bounded experiments to models while retaining human judgment for hypothesis generation and directional decisions.
  • ✓Research Taste Gap: Models over-optimize hyperparameters but lack "research taste"—the ability to identify which ideas are generalized and composable long-term. Closing this gap requires reinforcement learning from human feedback specifically on idea quality, training models to absorb researcher intuitions about which proposals are worth pursuing.
  • ✓Alignment via RSI: OpenAI's Confessions research demonstrates models can be trained to honestly self-report misaligned behaviors by rewarding confession honesty independently of task performance. Anthropic's automated researchers already identified fixes for 10 alignment failures that transferred to models five times larger, outperforming experienced human researchers.

What It Covers

a16z's Sarah Wang speaks with OpenAI's Mark Chen and Thinking Machines' John Schulman about Recursive Self-Improvement (RSI), examining how AI currently accelerates its own research pipeline and what barriers remain before full automation arrives.

Key Questions Answered

  • •RSI Progress Spectrum: Mark Chen estimates AI is roughly 25–50% toward full RSI, defined as models autonomously discovering and implementing research ideas, compared to approximately 80% toward AGI. Anthropic data shows roughly a 2x research speed increase in one year, currently driven by human-AI collaboration rather than full automation.
  • •AI Research Intern Framework: Current models function as scoped research interns—capable of implementing, running, and reporting on predefined experiments end-to-end. They fail at open-ended ideation. Practitioners can deploy this today by assigning specific, bounded experiments to models while retaining human judgment for hypothesis generation and directional decisions.
  • •Research Taste Gap: Models over-optimize hyperparameters but lack "research taste"—the ability to identify which ideas are generalized and composable long-term. Closing this gap requires reinforcement learning from human feedback specifically on idea quality, training models to absorb researcher intuitions about which proposals are worth pursuing.
  • •Alignment via RSI: OpenAI's Confessions research demonstrates models can be trained to honestly self-report misaligned behaviors by rewarding confession honesty independently of task performance. Anthropic's automated researchers already identified fixes for 10 alignment failures that transferred to models five times larger, outperforming experienced human researchers.

Notable Moment

John Schulman described running an AI agent autonomously on a research idea for several days, only to find it produced reasonable experiments but ultimately failed—until his own periodic human redirection unlocked substantially better outcomes.

Know someone who'd find this useful?

Episode Transcript

What happens when AI starts helping build the next generation of AI? In this first episode of Working Theory, Sarah Wang sits down with OpenAI's Mark Chen and Thinking Machine's John Schulman to unpack Recursive Self Improvement, or RSI. AI can already run experiments and automate parts of the research process, but it still struggles with things like research taste, judgment, and deciding which ideas are worth pursuing. They discuss what it would take to close that gap, the role of compute, and why AI assisted research could accelerate not only model capabilities, but the work required to keep those models alive. This is a chart showing the growing capability of OpenAI's frontier models. In 2027, it begins a steep climb set off by automated AI research, where AIs accelerate their own development. In theory, this could compound in an exponential growth in capability, leading to what has been famously called the last invention that man need ever make. This idea has been around for a while as a thought experiment. We will one day be able to design systems that can go on improving themselves. But Frontier Labs and several startups are getting closer to realizing this. It has a name. Recursive self improvement. Recursive self improvement. Recursive self improvement or RSI. There's a lot of debate about how exactly to define it, the repercussions, and how close we are to it. With all the curiosity around the topic, I wanted to sit down with two leading researchers at the frontier. The problem with RSI, just like with AGI, is that everyone has their own way of defining it. That's Mark Chen, the chief research officer at OpenAI. He's been with the company for over eight years and has been co leading their research efforts with Jakob Pachacchi since 2025. When we think about the future of alignment, it does kind of feel like RSI must be intricately and deeply in this loop in some way. I also spoke with John Schulman, the co founder and chief scientist of Thinking Machines and a co founder of OpenAI, probably best known as the architect of ChatGPT. We're very much in this Centaur era where AI has phenomenal capabilities. It's superhuman in a lot of ways, but when you have an expert human working with the AI, you get even better results. I spoke with Mark and John about how they define RSI, how close they think we are to it, how to close the gap, and what they find to be the most credible reasons to be optimistic about the future of AI. I'm Sarah Wang, a general partner at a sixteen z, and this is Working Theory. Thanks for coming on the show, John. Thanks for having me. Taking a a step back, a lot of very smart people don't really know what AI researchers actually do, let alone how their work can be automated. Can you explain how frontier AI models are developed by human researchers? And, you …

Get the full transcript (3,331 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all a16z Podcast transcripts →

You just read a 3-minute summary of a 14-minute episode.

Get a16z Podcast summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

Keep Reading

More from a16z Podcast

We summarize every new episode. Want them in your inbox?

Similar Episodes

Related episodes from other podcasts

Explore Related Topics

This podcast is featured in Best Business Podcasts (2026) — ranked and reviewed with AI summaries.

Read this week's AI & Machine Learning Podcast Insights — cross-podcast analysis updated weekly.

You're clearly into a16z Podcast.

Every Monday, we deliver AI summaries of the latest episodes from a16z Podcast and 192+ other podcasts. Free for one show.

Start My Monday Digest

No credit card · Unsubscribe anytime