> Ask a question one way, you get one answer. Ask it another, you trigger a different heuristic. You’d call that “intuition” if it came from a person.
No, This is the person who aces certain types of exams but doesn't know how to apply any of that "knowledge" in the real world because it was built upon rote memorization and not a true understanding. They're not looked upon favorably and I'd quietly question their intelligence.
> A human chef could have told you the exact same wrong thing with the same misplaced confidence
You're missing the point. The problem is that the answer changes depending on how the question is presented.
Chefs are indeed often mistaken in their understanding of how salt and yeast interact and how vulnerable yeast is to being killed by salt. The difference is that their answer doesn't flip like this. Their answer is fundamentally based upon an underlying understanding (however wrong), it's not simply a statistical continuation sequence of the words in my question.
This has serious implications because when a chef's understanding is corrected, they will update their underlying understanding and automatically apply that new knowledge in every scenario that relies on this understanding, where as an LLM can't do that, it can never be corrected without rewriting every instance of the training data where this falsehood comes from.
> You ask “where” and “by whom” the Turing test was passed, as if it’s some formal event with a ribbon cutting
No, I expect a research paper where the test design and its results are documented.
> The tools are shipping code
Yes I've seen the code they "ship", that's why your grandiose claims ring hollow.
No, This is the person who aces certain types of exams but doesn't know how to apply any of that "knowledge" in the real world because it was built upon rote memorization and not a true understanding. They're not looked upon favorably and I'd quietly question their intelligence.
> A human chef could have told you the exact same wrong thing with the same misplaced confidence
You're missing the point. The problem is that the answer changes depending on how the question is presented.
Chefs are indeed often mistaken in their understanding of how salt and yeast interact and how vulnerable yeast is to being killed by salt. The difference is that their answer doesn't flip like this. Their answer is fundamentally based upon an underlying understanding (however wrong), it's not simply a statistical continuation sequence of the words in my question.
This has serious implications because when a chef's understanding is corrected, they will update their underlying understanding and automatically apply that new knowledge in every scenario that relies on this understanding, where as an LLM can't do that, it can never be corrected without rewriting every instance of the training data where this falsehood comes from.
> You ask “where” and “by whom” the Turing test was passed, as if it’s some formal event with a ribbon cutting
No, I expect a research paper where the test design and its results are documented.
> The tools are shipping code
Yes I've seen the code they "ship", that's why your grandiose claims ring hollow.