Oh man...I most definitely want this! I already use ChatGPT to do most of my web searching now. My search engine use has probably gone down by around 75% in the last year.
This seems to me the logical next step for Internet navigation, especially now that results from search engines are getting worse due to spam. Even if search engines weren't getting worse, browsing would still go the way of generative AI, and I'm all for it.
Incidentally, I may have misunderstood something about generative AI (something that I hope is true). I was worried that biases and disinformation online would disrupt the truthfulness of the LLM AIs, but it seems that they may be robust to the nonsense because they read everything. Why are experts better at sorting the likely true bits from the likely false? Are they simply better-read than the rest of us in their particular areas?
Maybe future LLM AIs (perhaps paired with some explicit ability to form models and check for internal inconsistencies...if those features don't already organically emerge in LLM training) will become true experts by virtue of being so well-read.