Skip to content
Quotes.

Scott Alexander quote, Sep 24, 2026

Author, Astral Codex Ten · Sep 24, 2026 · Post, The Specter Of Neuralese

An AI with more real layers can actually be “smarter” in the conventional sense of the term.

Primary source

Words found on the source page on Sep 25, 2026.

More from Scott Alexander

Quote
For 100 steps in a row, the AI can shoot delicate subtle ideas from layer to layer at the speed of light. Then there’s one step where it has to encode them into twenty-six glyphs invented by Phoenician turquoise miners in 1800 BC. Then it has to re-encode the Phoenician glyphs into delicate subtle lightspeed ideas before it can do anything else.Sep 24, 2026· Post
The science of reading these thoughts is a subfield of AI interpretability, which is still in its infancy.Sep 24, 2026· Post
This story of misalignment says that LLM alignment makes AIs more aligned, RLVR makes them less aligned, and the exact level of alignment depends on how these two things interact or cancel out. But if this work generalizes, all the bad effects from RLVR get sequestered to RLVR like problems.Sep 23, 2026· Post
This trains the AIs to be focused on task success, which naturally risks including things like reward-hacking, cheating, and single-minded pursuit of stated goals at the expense of ethical injunctions.Sep 23, 2026· Post

On the same topics

Quote
It is important to not confuse massive steps in inference-time scaling, a dynamic which should be fairly predictable, with being the outputs of RSI, which is highly uncertain.Nathan LambertSep 19, 2026· Post
If you have five hours, you’re probably going to do a lot better. The AI models are pretty similar.Noam BrownSep 17, 2026· Interview
So I was like, “I don’t think we’re going to get it in 2026, probably not in 2027, maybe in 2028.” So it did happen a lot faster than I expected.Noam BrownSep 17, 2026· Interview
The concerning scenario is that, especially as these models are becoming more capable, we make them what we think is aligned, and they’re 99.9% aligned. Then we use these models to help us with the next generation of models, and they end up being 99.8% aligned. Then with each subsequent generation, we see an increasing degradation in alignment.Noam BrownSep 17, 2026· Interview
If I was in charge of an alignment or interpretability team I would never run out of experiments to run and things to try, including having the time to properly talk to the models in the first place.Zvi MowshowitzSep 14, 2026· Post
Over the summer there was a step change, as Mythos and Astra started kicking off the early stages of recursive self-improvement (RSI).Zvi MowshowitzSep 14, 2026· Post
About this quote
Said by
Scott Alexander, Author, Astral Codex Ten
Where
The Specter Of Neuralese
Topics
Research
Source
astralcodexten.com/p/the-specter-of-neuralese

A quote is one moment, not the whole argument. Read it in context at the source.

Sources: the primary source linked on each quote, checked weekly. A quote is one moment; read the source. Logos via logo.dev; trademarks belong to their owners.

New quotes by email

Fridays, only in weeks with new on-the-record quotes from data and AI leaders.

Double opt-in. Unsubscribe any time.