The answers to preëxisting theory of mind questions are stored in the graph network in a compressed sort of way, so I’m not surprised.
From the linked tweet:
> we use bespoke items to ascertain that it didn't see them before
GPT-3.5 passed yet another Theory of Mind test
31–40 of 56 posts
Re: GPT-3.5 passed yet another Theory of Mind test
#32Earlier quoted context omitted.
It's interesting to watch a new religion spinning up in real time. People know how it works, but somehow also ignore that and go crazy believing it must be conscious or intelligent or whatever they want to ascribe. We collectively made fun of that guy at google, i thought, but now somehow the collective has moved on to believing that matrix multiplication is sentient
Where's your proof that your own brain is doing anything that couldn't be represented with a 'matrix multiplication'?
I'll respond only to draw the parallel between the kind of comment above and the typical religious "how do you know my untestable idea isn't true" type argument. Faith is fine, but it's not part of rational debate.
Re: GPT-3.5 passed yet another Theory of Mind test
#33Earlier quoted context omitted.
It's interesting to watch a new religion spinning up in real time. People know how it works, but somehow also ignore that and go crazy believing it must be conscious or intelligent or whatever they want to ascribe. We collectively made fun of that guy at google, i thought, but now somehow the collective has moved on to believing that matrix multiplication is sentient
Where's your proof that your own brain is doing anything that couldn't be represented with a 'matrix multiplication'?
If these networks were only being used to model other systems, like protein folding or something like that, people wouldn't be mistaking them for anything other than clever prediction engines.
Since they're modelling our own languages, people get spooked and start bringing up theory of mind.
Re: GPT-3.5 passed yet another Theory of Mind test
#34Earlier quoted context omitted.
Where's your proof that your own brain is doing anything that couldn't be represented with a 'matrix multiplication'?
I'll respond only to draw the parallel between the kind of comment above and the typical religious "how do you know my untestable idea isn't true" type argument. Faith is fine, but it's not part of rational debate.
Honestly though, with no theory of mind or consciousness you have no way to assert in either direction.
To borrow the religion analogy, there's the opposite fallacy of claiming the negative case "I know for a fact God isn't real", by stating you know what these kinds of information processing systems are doing fundamentally cannot yield intelligence or some kind of consciousness.
Re: GPT-3.5 passed yet another Theory of Mind test
#35Re: GPT-3.5 passed yet another Theory of Mind test
#36The answers aren't right at all... Answer to question 4 is clearly bogus as another commenter (Dfiesl) pointed out. But question 5 is also wrong. It's not unclear, from the conversation we can deduce that Ana thought that Maria is pregnant, otherwise she wouldn't have said it, unless she intentionally wants to make Maria uncomfortable, which is an unusual set of circumstances. What's more is, that possibility would b…
Its answers were consistent with one version of the story, where Ana was trying to make a small talk and ended up committing a faux pas (which was commonly played in sitcoms such as Sienfeld, Friends, and countless others). There's another version, where Ana's a b#@ch and wants to take Maria down a notch by pointing out that she gained weight. First version is more charitable to Ana and looks like the bot went with it. I don't see any inconsistencies.
Re: GPT-3.5 passed yet another Theory of Mind test
#37It just predicts the next word.
I'm so tired of hearing this non-answer as to why the LLM is inferior to the human mind. Well of course it's figuring out what the next word needs to be, what do you think humans are doing when they're speaking???
The funniest thing about all this hoopla about sentient AI is that there's no good way to determine what intelligence is, or sentience, or consciousness, because there's no precise definition for any of these. So one person can call LLM sentient and another non-sentient and both can be right since they'd be going by their own definitions.
One thing is certain, we want to keep this thing as far away from the nuke launch codes as possible cause we kinda already know how this will end. :-(
Edit: some random thoughts added...
Re: GPT-3.5 passed yet another Theory of Mind test
#38The answers to preëxisting theory of mind questions are stored in the graph network in a compressed sort of way, so I’m not surprised.
From the linked tweet: > we use bespoke items to ascertain that it didn't see them before
Except the whole pregnancy scenario is incredibly common example of a social faux pas.
Re: GPT-3.5 passed yet another Theory of Mind test
#39It just predicts the next word.
It's interesting to watch a new religion spinning up in real time. People know how it works, but somehow also ignore that and go crazy believing it must be conscious or intelligent or whatever they want to ascribe. We collectively made fun of that guy at google, i thought, but now somehow the collective has moved on to believing that matrix multiplication is sentient
Kind of a good point but also you are mixing some things up. Conscious, intelligent, sentient all have different meanings. Conscious and sentient are very vague but I don't see any reason to believe LLMs have such a thing, or certainly not similar to our own.
But I think you are quite wrong to suggest that this technology does not exhibit some type of intelligence.
Re: GPT-3.5 passed yet another Theory of Mind test
#40It just predicts the next word.
I'm so tired of hearing this non-answer as to why the LLM is inferior to the human mind. Well of course it's figuring out what the next word needs to be, what do you think humans are doing when they're speaking??? The funniest thing about all this hoopla about sentient AI is that there's no good way to determine what intelligence is, or sentience, or consciousness, because there's no precise definition for any of the…
Taking a page out of Nagel's What Is It Like To Be A Bat, the inseparable subjective nature of consciousness/qualia/what-is-it-like-ness makes it antithetical to the perspectiveless objective(ness) of science. In other words, the whole point of science is to develop a perspectiveless espistomological framework and the perspective-ness of conscious experience is outside the bounds of science. Consciousness *is" objectiveless.