factual
GPT-2 Record Results Across Linguistics Benchmarks
A single generic GPT-2 model reached state-of-the-art results on 18 of 20 leading computational linguistics benchmarks, which Christian treats as evidence of progress toward AGI.
factualpending
Speaker
BrianEvidence Quote
“a single model was able to get record-breaking performance on 18 out of those 20 benchmarks. So that is the kind of thing that gives me some evidence that we're inching towards something like AGI.”
Created: 11/19/2026, 9:46:21 AM
Your notes
Loading notes...