I would imagine that citing sources would look more like finding sources that have a minimum distance to the output in a latent space representation? I don’t think it’ll make sense to track how each connection weight changed over time due to various inputs.

perplexity.ai/ specifically uses web sources to construct natural language.

I believe it's going to be in Bing soon?

> perplexity.ai/ specifically uses web sources to construct natural language.

I am going to guess that even though it uses the GPT language model, it works fundamentally differently than ChatGPT.

My guess is, Perplexity AI is a traditional search engine, that first finds a list of links that match your questions. And then it feeds text from those links to the GPT language model to write a summary of the results.

Whereas ChatGPT produces text based only on the trained deep neural net.

Maybe I am guessing wrong how Perplexity AI works. Maybe it works in a different manner.

It still has the exact same issues. I asked it to tell me about “Microsoft layoffs before 2022” and it confidently stated that Microsoft had no layoffs before 2022, with one of its sources even talking about the previous layoffs in 2009.

I’d have thought the short lived Galatica from Meta would have demonstrated more capability on this need given it was explicitly positioned as being to help Researchers.

I wonder if this is a special case for Explainable Models [1] — once you select an answer, another model helps describe and source the output.

[1] https://en.m.wikipedia.org/wiki/Explainable_artificial_intel...