This is why I think AI is a small bubble. It's sampling from training data and has trouble with things that are beyond the training data. The problem is that useful work is usually outside of this boundary of what currently exists. The usefulness of intelligence lies in finding outliers, i.e. novel solutions never seen before. That said, there is still a lot of use for LLMs like revolutionizing search and automating a lot of repetitive tasks that have to do with natural language.
So it's a big deal, especially if it can be combined with robots which might be the real revolution, but as far as intelligence, we don't yet have critical thinking machines.
This is why I think AI is a small bubble. It's sampling from training data and has trouble with things that are beyond the training data. The problem is that useful work is usually outside of this boundary of what currently exists. The usefulness of intelligence lies in finding outliers, i.e. novel solutions never seen before. That said, there is still a lot of use for LLMs like revolutionizing search and automating a lot of repetitive tasks that have to do with natural language.
So it's a big deal, especially if it can be combined with robots which might be the real revolution, but as far as intelligence, we don't yet have critical thinking machines.