AI Futures Project forecasters disagree significantly over the differing progress of SAR and TED-AI, while their superhuman AI timelines have been updated following METR’s latest report. METR estimates that Anthropic models are improving at twice the previous rate, using AECI as an internal metric. AIFP defines SAR as achieving capability parity with human researchers and TED-AI as consistently matching or surpassing the performance of the best human experts.
Source:LessWrong · Read original article
Leave a Reply