Alles wat Jan bezighoudt, interesseert en irriteert... en ook een beetje onzin...

vrijdag, april 12, 2024

Testing Perplexity AI


On of the problems with Large Language Models, or as the kids say "AI", is that these models tend to hallucinate but present their information with a certain resoluteness or self assuredness that can be mistaken for actually knowing the facts. 

The team behind Perplexity AI tried to tackle this by forcing the LLM to include copious amounts of links to the sources of the quilted information. Kagi's FastGPT is following a similar solution. 

I was pleasantly surprised when I tried this query:

What is deionized water?