Comment on UCSD: Large Language Models Pass the Turing TestparentComments−rfoo1yThe point is, it isn't marginally better. I agree the setup is not a demonstration of intelligence, but the difference is pretty significant. Not to mention that on conventional benchmarks Llama 405B is usually worse than GPT-4o.
Comments
The point is, it isn't marginally better. I agree the setup is not a demonstration of intelligence, but the difference is pretty significant. Not to mention that on conventional benchmarks Llama 405B is usually worse than GPT-4o.