Misaligned AIs could use killer robots to take over

Discussions tend to fall into 'AI fans' highlighting pros: AI-assisted coding at superhuman speed, breakthroughs in long-standing math puzzles, AI-enabled advances in science like new drug discovery, etc.

Then there's the critics: railing against datacenter buildouts, their energy consumption / environmental impact, job losses, geopolitics, open vs closed models, etc.

And of course new models, their performance, hardware required & so on.

But what about the risks? One can argue plausibility & probability of various scenarios. But risk of any such scenario unfolding hasn't disappeared, has it? And these risks are only increasing as AI-powered systems are given more eyes, ears, hands & feet, while taking humans out of the loop.

I wouldn't count myself hard into any of above camps. Overall AI is a mixed bag of enormous potential, pros, cons, limitations, uncertainties & unpredictability all around. But risks are real & worthy of discussion @ length, imho.

(btw love that AI 2027 site)