> In fact, it is easy to see how bad LLM AIs really are, just ask it a very narrow-focus question about a subject within your own area of expertise! 

For once, I find myself agreeing 100% withthe OP.  

In fact you don't need to be an expert to start seeing what he reports.  Just ask enough easy questions or offer a paper to be read.

In one case, I asked Perplexity about a presentation that was in in a Flipbook.  The 'bot kept writing about the container -- and read me from the Flipbbook manual -- not the contents.  It took a lot of repetitive work on my part to make it see that it could not see.

What is amusing is how the 'bot apologises when caught out and humbly thanks the operator for educating it.  I can see how some people could easily fall in love.

After a while the enchantment with this new toy wears off, after you see that it is cage-bound, wears blinders, is drugged, and can intuit what it is fed.  

We all know that what something is fed or allowed to eat determines what comes out the other end.  On some topics, it seems the 'bots are wise, on others, GIGO. 

Anyhow, a conversationwith a 'bot is never dull, and about two levels above what we usually experience on the 'net.  

Other than here, of course.

             ***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software.  For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html