ChatGPT will tell you the truth after it stops mattering

Fascinating read. I was almost tempted to use AI to summarize this (ran quite long) but glad I didn’t.

Ehh I get the point, but also:

* people want stable answers from AI, which is pretty much regression to the mean

* there is a safety element in there, not agreeing with the user every time is easiest if large institutions are leant on, tempering the sycophantic behaviour may result in this kind of behaviour

I think the thing that needs a finger on it is really of trust, and the anthropomorphic appearance of AI versus, say, a newspaper clipping, billboard, or ad. AI with this “human like” appearance draws

* unearned trust from people who expect social queues correlate with humanity

* unearned skepticism from people who see this gap between social queues and actual “care” as totally cynical control and manipulation

We are also in this crazy post-free-money world, where everyone was exposed to almost free, quality tools through the 2010s. Call it democratisation for a spell. This article feels like part of the hangover from that. In the past, institutions had exactly this sort of power.

What mattered yesterday is not going to be what matters tomorrow. Support your local library! Build your own!