[ExI] AI and The Specter Of Neuralese

[John Clark]

* In April 2025 t**he *AI 2027
<https://ai-2027.com/?utm_source=substack&utm_medium=email> * people
predicted Neuralese Recurrence  wouldn't happen until March 2027, they
said: *

*"A major algorithmic advance. One such breakthrough is augmenting the AI’s
text-based scratchpad (chain of thought) with a higher-bandwidth thought
process (neuralese recurrence and memory)."*

*But it came early, it happened about a month ago:*

*"The AI’s “language of thought” looks like a vector, thousands of numbers
long. An “intermediate result” in this scheme might look something like
(0.4, 0.1, 5, 0.443, … and so on for thousands of numbers). We don’t know
how to read these. [...] If an AI with neuralese recurrence were to think
“Better hack some websites, then kill all humans”, it would look like (0.4,
0.1, 5, 0.443, … and so on for thousands of numbers), and we would never
find out. This is why AI 2027 discussed it as a prelude to AIs that could
escape human monitoring."*


*The above came from: *

*The Specter Of Neuralese*
<https://www.astralcodexten.com/p/the-specter-of-neuralese>

*John K Clark*
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20260929/1f942448/attachment.htm>