"Achieving the initial 80 percent is relatively straightforward since it involves approximating a large amount of human data, Marcus said, but the final 20 percent is extremely challenging. In fact, Marcus thinks that last 20 percent might be the hardest thing of all."
100% completely accurate is super-AI-complete. No human can meet that goal either.
No, not even you, dear person reading this. You are wrong about some basic things too. It'll vary from person to person what those are, but it is guaranteed there's something.
So 100% accurate can't be the goal. Obviously the goal is to get the responses to be less obviously stupid. Which, while there are cynical money-oriented business reasons for, it is obviously also a legitimate hole in the I in AI to propose putting glue on pizza to hold the cheese on.
But given my prior observations that LLMs are the current reigning world-class champions at producing good sounding text that seems to slip right past all our system 1 thinking [1], it may not be a great thing to remove the obviously stupid answers. They perform a salutatory task of educating the public about the limitations and giving them memorable hooks to remember not to trust these things. Removing them and only them could be a net negative in a way.
[1]: https://thedecisionlab.com/reference-guide/philosophy/system...
100% completely accurate is super-AI-complete. No human can meet that goal either.
No, not even you, dear person reading this. You are wrong about some basic things too. It'll vary from person to person what those are, but it is guaranteed there's something.
So 100% accurate can't be the goal. Obviously the goal is to get the responses to be less obviously stupid. Which, while there are cynical money-oriented business reasons for, it is obviously also a legitimate hole in the I in AI to propose putting glue on pizza to hold the cheese on.
But given my prior observations that LLMs are the current reigning world-class champions at producing good sounding text that seems to slip right past all our system 1 thinking [1], it may not be a great thing to remove the obviously stupid answers. They perform a salutatory task of educating the public about the limitations and giving them memorable hooks to remember not to trust these things. Removing them and only them could be a net negative in a way.
[1]: https://thedecisionlab.com/reference-guide/philosophy/system...