The Inevitable Failure of Developing AI, so-called, via LLMs.
Testable Predictions.

Sentient, never mind hostile, AI is not even on the horizon yet. Fire is much easier to create, and can harm much sooner than any AI. etc. etc.

This is "thinking about interesting concepts" - not engineering.

The argument is based on some absurd premises. Please don't point out that we don't have such AIs yet. But that hasn't stopped the AI Doomers from fretting about Skynet, either. So, cut me some slack.

@edzitron.com - for your consideration.

I should have used the pronoun 'they' in the first place. Turns out he identified as male, according to his journal. That should settle it.

I didn't know that. So, he left a note. That settles it.

If there ever was a very strong case to be made for using "they" as the pronoun, this is probably it.

This raises an interesting conundrum: should they be referred to by their gender assigned at birth or their performed gender? Both options are very problematic.

Unless she left notes on how she identified, it's most likely she disguised as a man. I doubt her social context provided her with the concepts necessary for considering herself potentially trans-gender.
It's difficult to think thoughts in isolation from your cultural context.

There's a valid debate to be had about the implications of such AI in principle, but there is little point in worrying about it at the expense of more pressing issues (wealth and income inequality, the environmental crisis, war...).
Fix the problems we have, then worry about problems we may have.

Unless you live south of the equator :)

absolutely.