There is a lot of work on systems that use transformer models that will get closer to “the truth”. For instance you could run a query through a normal search engine, then use a much more expensive cross-encoder to screen the top results, then have the transformer summarize the documents. In general transformers could be joined to other specialist systems to answer questions in domains like math.

Truth, however, is the most problematic idea in philosophy, and simply introducing the concept of ‘truth’ erodes the truth. (E.g. makes possible the statement “I am lying now” for the logical and for the X Files to assert “the truth is out there”)

The most interesting question though is how people have reacted to ChatGPT. I mean, Wolfram Alpha is a great question answering system but it didn’t cause a panic for Google and Bing. In our strange times people might rather have a bullshit engine than a search engine, and under the sign of “the innovators dilemma”, search engines might rush to disrupt them own business as a result.