New study finds: AI still relatively stupid
Meta has a new LLM benchmark focusing on things people are good at but AI systems still find hard. Researchers show that human respondents obtain 92% vs. 15% for GPT-4 equipped with plugins.
Meta has a new LLM benchmark focusing on things people are good at but AI systems still find hard. Researchers show that human respondents obtain 92% vs. 15% for GPT-4 equipped with plugins.