[WarOnPrivacy]

I'd fix AI' inherent untrustworthiness.

I'd make AI unable to resist presenting the knowledge is has; I'd fix AI so it would fully answer the question I asked.

I'd make precursory fluff and post content review and explanation - opt-in only.

I'd fix AI' gaslighting-like inability to correctly parse an inquiry.

I'd fix AI' inability to learn from the mistake it just made.

I'd fix the near ubiquitous bad choice to put unasked-for AI up front. If it's not requested, it's not wanted.

[zyruh]

This is great - thank you!