Toggle light / dark theme

Recursive self-improvement has started

It s so close now, RSI is the key. RSI could be solved by next summer, if it is, We could have ASI by end of 2027.


If you asked me earlier this year how far away we are from superintelligence, I’d probably have said at least the better part of a decade.

But over the last three months, from debating people on and off the podcast, I’ve updated my views a ton.

Let me walk you through how my thinking has changed.

If you want to read the written version (with footnotes and links) you can check it out here: https://www.dwarkesh.com/p/rsi.

Music by ‪@tomfoxcatalog‬

0:00:00 – AI is already speeding up AI progress.
0:02:03 – Math is a preview of what will happen to AI research.
0:11:10 – Automating the best researchers would speed up AI progress a lot.
0:14:24 – We might get a singularity or a spark.
0:17:21 – Superintelligence will be a really good meta-learner.
0:25:59 – AI R&D is largely verifiable.
0:33:59 – Objection: continual learning.
0:36:44 – Objection: finding the right verifiable proxies.
0:38:26 – Objection: data.
0:39:58 – Objection: experiment compute.

Leave a Comment

Lifeboat Foundation respects your privacy! Your email address will not be published.