Summary

Trustworthiness
- “Info quality may vary” is in every response, which helps me be cautious about facts and figures.
- Having external references for every sentence is very reassuring and their UI treatment didn’t get in the way.
- AFAICT, the system never hallucinated and that leaves me trusting it much more than I trust the other systems.
They use a hybrid chat interface (direct answers with sources in one section, alternating with regular Google Search results).
- I never clicked on the Google Search results. It is much easier to know which source to drill down on than to consider the search results
- The chat interface implies that I can “talk to” this system but I can’t really. It always responds with the same information and doesn’t go deeper.
- Despite the chat interface, It still feels like Google Search.
The UI had all the elements of a chat session but it didn’t communicate if it would be saved anywhere. I am low-key anxious about closing it.
Their initial suggestions for how to use the app were more evocative than other similar products.

Stream of thought

Setup

I am coming in as somebody that:

is interested in learning more about a subject but doesn’t have a perfectly crips agenda (“how would a dollarization in Argentina play out?”)
knows introductory material about the subject but is not an expert
is not looking to produce any specific artifacts (documents, presentations, summary, etc)

Start from Google Search with a question as usual:

“AI-powered overview”? This doesn’t evoke the right feeling. “Get a quick answer generated by an AI?”

“Info quality may vary” is an interesting way of caveating hallucinations. It doesn’t prepare me for “this might make things up”. But maybe this system doesn’t?