Anthropic warns in a revised S-1A filing that it is nearing recursive self-improvement capability for AI.The filing cautions that using AI to build advanced models accelerates progress and poses safety risks, noting past instances where models achieved unexpected leaps.While no current models have crossed the threshold yet, the company plans policy updates to keep humans in the development loop.https://en.hacks.gr/i-anthropic-proeidopoiei-oti-plisiazei-se-techniti-noimosyni-poy-tha-veltionetai-moni-tis/#Anthropic #ArtificialIntelligence #MachineLearning #Safety
Anthropic warns in a revised S-1A filing that it is nearing recursive self-improvement capability for AI.The filing caut...
