AI (or what we call AI currently) is definitely capable of making decision it wasn’t trained for, it’s the whole reason why LLM are so effective, they learned enough patterns about how humans think and react, that they can apply it to other situations.

Now, if you talk about RELIABLY making decisions then I definitely agree with you. They can’t even be reliable with stuff they extensively saw in training, when you get out of the “confort zone” the odds of it taking the correct decision drops fast.

But with AI companies trying to get as much data as they can from their workers and from common people, and with the scary new trend of smart glasses (even though I’m a fan of augmented reality as a concept) we will see how things evolve…