This discussion ignores the basic design of LLMs --- they are fundamentally statistical and probablistic. I don't have a really good definition for "intelligence" but it seems rather obvious that it is not achievable by simply rolling dice or building a really large database of the internet.
In conclusion as long as LLM performance continues to improve we aren’t in an AI bubble.
See the response to ChatGPT 5:
https://www.analyticsinsight.net/chatgpt/why-chatgpt-5-didnt...
This discussion ignores the basic design of LLMs --- they are fundamentally statistical and probablistic. I don't have a really good definition for "intelligence" but it seems rather obvious that it is not achievable by simply rolling dice or building a really large database of the internet.
[dead]