Uploaded September 2024 | Updated November 2026, 1 day ago
GPT-4 has a 3% hallucination rate, while Google’s PaLM 2 nears 27%.
Even with billions poured into their development, the best LLMs still struggle with contextual comprehension.
Ask how to make pizza, and it might suggest using glue to stick the cheese.
Why?
Because it lacks real-world experience and context.
It can write unit tests but doesn’t grasp the broader goal of ensuring code quality.
Because its responses are only as good as the datasets it’s been trained on.
We discuss this and more in EP 02 of Floating Points, featuring Mike Hamilton, CIO of Cloudflare.
#pizza #ai #tech #chatgpt
GPT-4 has a 3% hallucination rate, while Google’s PaLM 2 nears 27%.
Even with billions poured into their development, the best LLMs still struggle with contextual comprehension.
Ask how to make pizza, and it might suggest using glue to stick the cheese.
Why?
Because it lacks real-world experience and context.
It can write unit tests but doesn’t grasp the broader goal of ensuring code quality.
Because its responses are only as good as the datasets it’s been trained on.
We discuss this and more in EP 02 of Floating Points, featuring Mike Hamilton, CIO of Cloudflare.
#pizza #ai #tech #chatgpt










