Reblogged by bcantrill ("Bryan Cantrill"):
cyberchris@discuss.systems ("Chris") wrote:
I really liked those examples of concrete, gnarly engineering challenges. Definitely puts things into perspective on what might an AI *actually* need to solve.
Your point regarding the "humanity" that AI lacks, is loosely referred to as "agency" in the AGI space. And yes, most of the popular AI tools (e.g. the usual LLMs like ChatGPT) have *zero* agency!
I don't know if the doomer twitter folks mention this distinction, but yeah, so far I haven't seen much convincing evidence of an agentic AI having anywhere near the ability needed to solve those engineering challenges autonomously.