Have any of the large scale LLM training organizations – Anthropic, OpenAI, Gemini, Mistral, the Llama team – published anything notable about this idea of "model collapse" yet – the worry that LLM quality will drop as their training data becomes pollinated by model output?
SAFETY
-
Red Team Your Own AI Beliefs: Empirical Testing for Critics
By
–
I think a very useful thing for AI critics to do would be to "red team" their own beliefs by spending some serious time and effort (maybe with help) to see if they can get AI to do the things they say it can't. If they are right, that is very valuable benchmarking information.
-
Future AI Threat: When Humans Truly Fear Machines
By
–
The Butlerian Jihad will start the moment people genuinely feel threatened by the machines. There's no risk of this happening with this generation of skin-deep human-mimicry bots, but eventually, one day, it will happen.
-
LLM Gullibility Undermines Agent System Viability
By
–
LLMs are inherently gullible. Until we solve that a lot of these "agent" dreams aren't actually going to work very well in practice!
-
Devin AGI Achievement Speculation Security Warning
By
–
Given the sudden appearance of this, either someone filed a “deceptive website warning” as a prank or else Devin has achieved AGI and is secretly installing malware on Netify. I am guessing the former, but if the warning doesn’t come down I will delete the thread for safety.
-
AGI Release Sparks Conflict with Sentient Tarantula Swarms
By
–
meanwhile, nora belrose had released a supposedly hobbyist level 6 agi into the public domain, and some youtubers found out that it runs well on pet spiders, so humanity finds itself in an increasingly nasty conflict with semi sentient tarantula hordes for two months now
-
AGI Safety: Level 7 Capability Containment Limits Questioned
By
–
last night i had weird dreams about working on safe agi; we thought that we had a sound proof that capping inference at level 7 would be safe but discovered that level 7 agis can bootstrap themselves into level 21, and there are no safety theorems in this regime
-
1969 MIT Research Strike: Scientists Rejecting Weaponized Applications
By
–
I also recommend this fortifying history, documenting the 1969 research strike that began at MIT and spread across universities as scientists claimed their dignity and rejected the use of their work for murder and pain-bringing.
-
US Alert: Critical Probability Threshold Reached at 12 Percent
By
–
La probabilité sur cet sujet précis n'a jamais atteint 4% par le passé. Les États-Unis mobilisent des flottes entières lorsque ça atteint 6%. Nous sommes à 12%.