News for hackers. Stuff that matters.
Re:An automatic theorem proving project by cubefox on Thursday April 28, 2022 @04:18PM (#31197687)
DeepMind's recently introduced 540 billion parameter language model PaLM can already explain complex jokes and and reason logically by explicitly describing reasoning steps. The results are quite impressive. https://ai.googleblog.com/2022/04/pathways-language-model-pa...

To be sure, this happened just two years after GPT-3. So 5 years for something which Gowers wants, but based on a large language model, seems in fact quite realistic.

There are no comments on this comment yet.