Diary 2025-01-07
kaityo256 Q: Why should I learn something that I can find on the internet? A: Otherwise, it would be impossible to judge the authenticity of what you find on the Internet.
AI is the same.
nishio This logic implicitly assumes that "learned humans are better judges of truth than AI", but it is quite doubtful that this is true even after the time required for a person who is not capable of judging now to acquire the ability to judge. (The fact that people who have already learned are more capable of judging is not relevant to this discussion.) hrjn I don't think it is necessary to judge the truth based only on the information, since there are surprisingly many things that can be experimentally judged as true or false. For example, we can tell if a program proposed by an AI is correct by running it. Geocentric theory can be confirmed by observing the stars. News can be checked for inconsistencies by consulting multiple sources.
I was wondering if the search is the same.
hrjn AI can be viewed as search 2.0, so to speak, which allows for more abstract information retrieval. It's simply this. As the level of abstraction increases, it is inevitable that irrelevant content is somehow included. However, the time and effort required to search for information is dramatically reduced, so it is easy to compare and contrast from various perspectives. Agents, for example.
hrjn It may be less efficient than search for a one-shot query investigation, but nowadays AI can throw queries with a high level of abstraction to the extent that it can throw multiple queries and comparisons, I think the key point is that the cost of the overall true/false decision is low as a result. ---
This page is auto-translated from /nishio/日記2025-01-07 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.