Anthropic co-founder Jack Clark says 60% chance of RSI by end of 2028:
@goodside
-
RLHF requires pairwise comparison, not thumbs-up/down
By
–
The thumbs-up/down buttons aren’t for RLHF. The HF in RLHF is selecting the better of two outputs for the same prompt (or, if you have more than two, ordering them best-to-worst). Pointwise data (thumbs, star ratings, etc.) generally doesn’t work well AFAIK.
-
Proposal to apply safety testing only to high-revenue labs
By
–
That honestly doesn’t sound hard at all; just make the onerous safety testing requirements only apply to labs above some threshold of revenue.
-
LLM perfectly explains tweet about prompt injection origins
By
–
Excerpt from a Claude 4.7 Research report; prompt: “Explain the origins of prompt injection.” Surreal to see an LLM perfectly explain a tweet I made specifically about text that tricked then-SoTA LLMs, accurate down to my use of doubled exclamation points:
-
AI’s evolving job market: Exciting new roles emerge, then disappear
By
–
AI will take some jobs, but it will create countless new jobs too—exciting jobs we can’t even imagine yet. A year later those will also be done by AI, but there will be new jobs—exciting jobs we can’t even imagine yet. Six months later those too will be done by AI, but
-
ChatGPT 5.5 Pro / Images 2.0 creates a D’ni numeral wall clock photo
By
–
ChatGPT 5.5 Pro / Images 2.0 generates a photo of a wall clock using D'ni numerals—the fictional base 5 numeral system from Riven: The Sequel to Myst (1997):
-
LLMs predict next token with goblins involved
By
–
All LLMs do is predict the next token and also goblins are involved not sure how exactly but I’m sure about the first part
-
ChatGPT 5.5 Pro generates chess PDF with quiescence search analysis and QR code
By
–
ChatGPT 5.5 Pro (Extended) creates a multi-page PDF report about a chess game it simulates to test its own reimplementation of quiescence search, with its analysis of questionable moves and a QR code for viewing the game in the lichess PGN viewer
-
ChatGPT’s cake text transcription silently fixed punctuation, author admits mistake
By
–
You're absolutely right. ChatGPT's transcription of the cake's text silently corrected missing punctuation on the three lines you noted. I'm leaving this post up as a funny demo but anyone considering citing this in serious work should know about this oversight on my part.
-
Pro for better image generation with complex reasoning
By
–
Image output is better with Pro in my experience for prompts where substantial reasoning is needed. E.g. here it computes the face colors via code before generating. It may well work for Thinking, but I used Pro here so I disclose that.