This is really interesting, considering that this particular vocabulary is basically free to use since it doesn't really describe anything real. I will still cringe at most of them, though.
One remark on your premises: LLMs don't display intelligence... I don't think this is true unless you go for a definition of intelligence that has assumptions about consciousness built-in as a requirement. If they don't display intelligence what is the fuss all about? You might also be able to say they are "not really intelligent", but they certainly do "display" something like intelligence. But now we are probably caught in the meaning of "display".
There are many definitions of intelligence - English Wikipedia lists "capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving". For me, a program that provides a pretty good solution to language modelling doesn't create any of these. But it perhaps does lead us to reflect A LOT on what we associate with linguistic fluency.
Hm, but are we in an all-or-nothing situation here? By that definition, we need to change the name of the field, since artificial _intelligence_ research has never produced anything that's intelligent. A bit of a downer if you ask me :)
Yup, I think that'd be a fine start - the rumour is it very narrowly lost out to "cybernetics". I'm open to new definitions of intelligence, but less open to using a human definition of intelligence to describe a very fancy and low-ppl language model, you know? Oh, another kind of vocabulary gap!
Depends on the population, but in this context, I'm not sure this would be a well-qualified poll - the sampling is done for good qualitative data, not quantitative, and there's definitely a tension there (assuming time spent interview is not infinite. which it definitely is not!)
I am hesitant to challenge you more here because I am having a hard time untangling what my beliefs are vs what I believe you believe and how that influences our respective answers :) And now add popular vs scientific understanding of intelligence to that and it's a mess.
I definitely had an emotional response to the "LLMs don't display intelligence" sentence along the lines "Oh, there's no way you really believe that, do you?".
I have seen a lot of reactions that I would think of as post-hoc rationalizations. What about the "gut-feeling-based definition of intelligence"?
This is really interesting, considering that this particular vocabulary is basically free to use since it doesn't really describe anything real. I will still cringe at most of them, though.
One remark on your premises: LLMs don't display intelligence... I don't think this is true unless you go for a definition of intelligence that has assumptions about consciousness built-in as a requirement. If they don't display intelligence what is the fuss all about? You might also be able to say they are "not really intelligent", but they certainly do "display" something like intelligence. But now we are probably caught in the meaning of "display".
There are many definitions of intelligence - English Wikipedia lists "capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving". For me, a program that provides a pretty good solution to language modelling doesn't create any of these. But it perhaps does lead us to reflect A LOT on what we associate with linguistic fluency.
Hm, but are we in an all-or-nothing situation here? By that definition, we need to change the name of the field, since artificial _intelligence_ research has never produced anything that's intelligent. A bit of a downer if you ask me :)
Yup, I think that'd be a fine start - the rumour is it very narrowly lost out to "cybernetics". I'm open to new definitions of intelligence, but less open to using a human definition of intelligence to describe a very fancy and low-ppl language model, you know? Oh, another kind of vocabulary gap!
If you polled people in your sample, do you believe if you asked "does the model display intelligence?", you would get a majority negative answer?
(I don't think that would happen.)
Depends on the population, but in this context, I'm not sure this would be a well-qualified poll - the sampling is done for good qualitative data, not quantitative, and there's definitely a tension there (assuming time spent interview is not infinite. which it definitely is not!)
I am hesitant to challenge you more here because I am having a hard time untangling what my beliefs are vs what I believe you believe and how that influences our respective answers :) And now add popular vs scientific understanding of intelligence to that and it's a mess.
I definitely had an emotional response to the "LLMs don't display intelligence" sentence along the lines "Oh, there's no way you really believe that, do you?".
I have seen a lot of reactions that I would think of as post-hoc rationalizations. What about the "gut-feeling-based definition of intelligence"?