@Miles_Brundage: Read the roon, folks
Summary
roon (@tszzl) 声称各前沿实验室的研究人员私下乃至公开都承认对齐问题尚未解决，若属实则急于推进RSI（递归自我改进）是史无前例的愚蠢之举，呼吁很快显著放缓前沿研究；Miles Brundage 转发并呼吁大家重视这一警告。
View Cached Full Text
Cached at: 10/13/26, 05:35 AM
Read the roon, folks
roon (@tszzl): the researchers at all the labs would tell you in private and sometimes in public that alignment has not been solved. if this is true, then racing into RSI is a folly that is unprecedented in world history. we will need to significantly slow the frontier of research very soon
Similar Articles
@beffjezos: Talked to Roon about this tweet privately, he does support open source he's just pessimistic about what could happen. M…
A Twitter discussion where Roon expresses pessimism about the future of open source AI, predicting bans after potential disasters, while others urge understanding.
Roon suggests AI advancement/capabilities has outpaced alignment initiatives
Screenshots of roon's posts suggesting that AI capability advancement has outpaced alignment progress, sparking discussion about the current state of AI safety efforts.
My 3 cents on RSI
Vadim Fedenko shares a technical analysis of Recursive Self-Improvement (RSI), arguing that true RSI requires improving capability faster than complexity and expanding architectural space rather than just optimizing within fixed parameters. He doubts recent claims by xAI and Anthropic that RSI could arrive within a year, citing LLMs' poor subtractive engineering skills and current reward functions that ignore complexity.
@TheAhmadOsman: Roon already told us their plans btw Whatever disasters we see in the future, it’s made by them to serve their narrativ…
A tweet thread discusses how AI companies might manufacture hazards to drive regulatory capture and control the narrative, citing regulatory strategies.
@Miles_Brundage: Can’t speak to the details here but I will say that 1. one generally does not hear the best things about the data indus…
A discussion highlighting poor practices in the data industry, specifically regarding RLVR training data for AI models, with widespread sloppiness noted.