[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

A.I. - Introducing the Backgammon Rules Assistant / Clarifications

Posted By: Timothy Chow
Date: Wednesday, 12 June 2024, at 11:55 a.m.

In Response To: A.I. - Introducing the Backgammon Rules Assistant / Clarifications (Sebalotek)

I want to point out a common misconception surrounding LLMs (Large Language Models, such as ChatGPT). LLMs represent a significant advance in AI, but what they excel at is language fluency. They do not represent a significant advance in accurate logical reasoning.

The distinction is important, because we humans have a strong tendency to assume that if someone, or something, can "talk a good game," then they must know what they're talking about. But LLMs are specifically trained to be able to talk a good game only, and not to reason logically. You should therefore trust an LLM about as much as you would trust your enemy's smooth, fast-talking lawyer. Maybe less.

Subtle, nuanced reasoning about borderline cases is precisely where LLMs are most likely to go astray. Even if the Backgammon Rules Assistant is not a pure LLM and is supplemented by other types of AI to guard against the most egregious types of "hallucinations," I don't think it will achieve its stated purpose, and it might even generate more heated arguments than if one were to simply rely on the old-fashioned method of looking up what the rules say.

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.