a16z Launches Working Theory With Mark Chen and John Schulman on Recursive Self Improvement
For decades, the idea of AI improving itself was just a thought experiment. Today, AI is already helping researchers at frontier labs run experiments, write code and test ideas faster. So how far are we from AI that can do the research on its own? In our first episode of Working Theory, a16z General Partner Sarah Wang explores recursive self-improvement with two leading AI researchers: Mark Chen, Chief Research Officer at OpenAI, and John Schulman, co-founder and Chief Scientist of Thinking Machines Lab, and co-founder of OpenAI. They cover what RSI actually means, why research taste is so hard to teach, and why AI may be one of our best tools for solving alignment. 00:00 The last invention 01:58 What is RSI? 03:21 How far are we? 05:56 Getting to full RSI 10:57 Aligning RSI 15:08 Working theory YouTube: t.co/FSHJdT8Lym @markchen90 @johnschulman2 @sarahdingwang
Context
Two senior researchers discuss how close AI is to doing AI research on its own, a question that shapes lab strategy and alignment.
a16z posted the first episode of its Working Theory series, hosted by general partner Sarah Wang, with Mark Chen, Chief Research Officer at OpenAI, and John Schulman, cofounder and Chief Scientist of Thinking Machines Lab and cofounder of OpenAI. The post says they cover what recursive self improvement means, why research taste is hard to teach, and why AI may help solve alignment.
Deep dive
a16z posted that for decades the idea of AI improving itself was a thought experiment, that AI is now helping researchers at frontier labs run experiments, write code, and test ideas faster, and that the open question is how far away AI is from doing research on its own.
In the first episode of Working Theory, a16z general partner Sarah Wang explores recursive self improvement, called RSI in the post, with two researchers. Mark Chen is Chief Research Officer at OpenAI. John Schulman is cofounder and Chief Scientist of Thinking Machines Lab and a cofounder of OpenAI. The post says they cover what RSI actually means, why research taste is so hard to teach, and why AI may be one of the best tools for solving alignment.
The post lists chapters with timestamps: The last invention at 00:00, What is RSI at 01:58, How far are we at 03:21, Getting to full RSI at 05:56, Aligning RSI at 10:57, and Working theory at 15:08. The video is about 16 minutes long.
The post is a promotion for a podcast episode and does not include quotes or findings from the conversation. The views expressed by the guests were not reviewed for this entry.
What to expect next
Viewers will look for any concrete claims from the guests about timelines and how much of AI research labs already automate. Later episodes of the series and reaction from other researchers will show whether the discussion shifts the debate over recursive self improvement. Viewers can check the chapter list in the post to jump to the sections on how far AI is from full recursive self improvement and on aligning it, which are the parts most likely to be quoted.
