@simon

The axis for me wasn't useful/useless.

It was danger/safety. Still is.

I'm a Systems Reliability Engineer by trade, and it leaks all over everything. I look not at successes, but at failure modes. And human beings are very capable of deceiving themselves about just how "smart" things that seem human are. We're lazy too and it's a scary mix.

We've already seen horror stories about lawyers submitting hallucinated cases to court. We've heard about LLMs suggesting people eat very poisonous mushrooms. We see people trusting self-driving with very grim results.

I don't want to throw the baby out with the bathwater - but I trust the output as far as I can throw it. In some cases you can easily check it - coding, for example. It might still be subtly wrong, but humans make subtly wrong mistakes too. In other cases - people might die.

A chainsaw is a very powerful and useful tool that occasionally kills someone. I don't have one. I just worry we're in the "reckless" phase of AI in general, and I hate to see people hurt.