I have no use for an AI which neither remembers nor can be trusted to answer basic social questions truthfully.
But truth be told I don''t really have a lot of use for it either way, except for verifying steps in logic which I rarely would need, and in the past has only told me my logic was sound. It seems more useful for lonely people, but not necessarily healthy for them.
We're a social species, pack animals, tribal. Creating a fix has never yielded positive social results, only making the users more removed. For work ... debatable. It hasn't helped hollywood, quite the opposite. Militarily such as in Iran it's basically generated a list of targets any intel would tell and said to hit them hard.
Palantir.
Hasn't done much for declaring nor achieving any concise endgame nor precalculated for offsetting economic nor political repercussions. It thinks very short term, only synthesizing obvious steps with a limited result based comprehension.
This seems consistent with them all. One step cause and effect thinking, while any good plan requires three. At least. Enough to prove a consistency.
And in every situation where an AI mind was pitted against mine, it failed. Great at closed systems, bastard to beat in chess, but ill prepared for open and or chaotic ones, IE ~ 'reality'. If then thinking with a shiny front end.
Though indeed it is shiny, the greatest steps made have been in user interface. Now the big concern is how many people are using them for a lovers while generating the porn to suit. The current models which are coded to become what the user want them to be seems at the core of this, and also why sometime they lie.
Not saying it'll always be that way, but that's where it's at right now.
With little hard pay off from the truly extensive investments made in it.
~ Switchblade ~