Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Large language models lack deep insights or a theory of mind

arxiv.org

Post reply on HN
Print view View original

241–250 of 270 posts

Re: Large language models lack deep insights or a theory of mind

#241

Post by creer »

mr_toad wrote: post #221
krainboltgreene wrote: post #167

Earlier quoted context omitted.

Look this is the only time I'll engage in this sort of discussion on HN[1], but first Donald Knuth is a real Human and it's extremely weird to position world class experts as something otherworldly. Second, suppose you got what you wished for (you used the "us" pronoun), is that not a sentient mind that you're forcing to do your labour? Does that not raise a ton of red flags in your ethics? [1] normally I find HN dis…

> is that not a sentient mind that you're forcing to do your labour? Does that not raise a ton of red flags in your ethics? Maybe and maybe not. Being sentient or even sapient doesn’t mean it has human emotions, feelings or motivations. You have to be very careful when ascribing naturally evolved emotions and motivations to something that is not a naturally evolved intellect.

Does it need "human emotions, feelings etc" to "deserve" protection and status? Human society answered no a long time ago. But still it will be interesting to see what protections are granted to entities that are unarguably also machines.

Re: Large language models lack deep insights or a theory of mind

#242

Post by creer »

tinco wrote: post #30

I think that if they would, that would be very surprising and indicative of a lot of wastefulness inside the model architecture. All these tests are simple single prompt experiments, so the LLM's get no chance to reason about their responses. They're just system 1 thinking, the equivalent of putting a gun to someone's head and asking them to solve a large division in 2 seconds. I bet a lot of these experiments would…

At the same time, if LLMs are based on all or enough human writings then don't they necessarily contain a theory of mind? A rather general, smoothed out and still neurotic one probably. But still just like an LLM can't be expected to have a specific knowledge of hydraulics, it also has read more hydraulics than even experts might be expected to. That's the entire issue about it, right? This issue of "is most of our mind basically just mixing and matching stuff we have seen, read, heard?" Do humans have some magical theory of mind that somehow stands ASIDE from all the "normal" learned stuff?

Of course, yes, we do know one thing that's missing in LLMs, which is "loop and helpers" like you describe. Which I'm sure many people are currently hacking at - one way being for the LLM to talk to itself.

But as for "a theory of mind", if enough writings served as input, then LLMs do have plenty of that.

Another question is whether LLMs are raised to behave like humans (which might be where they most NEED some theory of mind). Of course not. The ones we know most about are only question answerers. The theory of mind they might have (that is not negated by the lack of loop and internal deliberation) may be overwhelmed by the pre- and post-processing: "no sex, no murder plots, talk to the human like they are 5, bla bla bla". And yet you can ask things like "Tell it like you are speaking to 5 year olds who want to have a fun time". Some theory of mind makes it through.

Re: Large language models lack deep insights or a theory of mind

#245

Post by rf15 »

I work in the field. It's just not how text-token-based autoregressive models can ever work. I can't talk about my work of course, but even a quick glance on Wikipedia can tell you they'd need to be at least a symbolic hybrid, which is not being pursued(?) by the big players at the time.

Re: Large language models lack deep insights or a theory of mind

#246

Post by TeMPOraL »

somewhereoutth wrote: post #228
TeMPOraL wrote: post #208

Earlier quoted context omitted.

> What they've proven is that you can build functioning systems that perform intelligent tasks purely at the level of syntax. But there is no (or very little) understanding of semantics. That's not how LLMs work though, and I'm increasingly convinced that "syntax" and "semantics" are turning into annoyingly useless ideas people forget are descriptive , in the same way grammar books and dictionaries are descriptive .…

> big enough Not even slightly. See Cantor's diagonal argument. See also Plato's cave for syntax vs semantics.

> Not even slightly. See Cantor's diagonal argument.

Cantor's diagonal argument, if important, can just become one of the hundred thousand dimensions of the latent space. "Coordinate #12321 - degree of similarity to diagonal argument".

> See also Plato's cave for syntax vs semantics.

I don't see it. All there ever is are projections. There are just more dimensions in the mind space than in the physical space.

Re: Large language models lack deep insights or a theory of mind

#247

Post by TeMPOraL »

AnimalMuppet wrote: post #236
TeMPOraL wrote: post #197

Earlier quoted context omitted.

Yes, but as the saying goes, anything can fly if you strap a powerful enough engine to it... so we've demonstrated basic capability, and the rest is now optimizing its performance a couple orders of magnitude.

Not at all. "Anything can fly", but not everything can fly without killing you. The fundamental problem is three-axis control with stability. A source of power does exactly nothing to solve that.

No, that just means you need a bigger engine. (And, perhaps, an inertial dampener, to survive the gees.)

Think of it this way: the more thrust you get, the farther you'll go before control problems start to manifest. Keep pushing, and eventually the rest of the craft becomes a rounding error in the math :).

Re: Large language models lack deep insights or a theory of mind

#248

Post by nprateem »

corethree wrote: post #238
nprateem wrote: post #231

Earlier quoted context omitted.

> A rock is not conscious. That's obvious. But it's not obvious at all. It may possess consciousness in a way we can't relate to or communicate with. This is the whole problem with consciousness and has been discussed by philosophers for centuries. We each appear to be conscious but can't be certain anything else is or isn't.

You're not reading my argument. You responded to one part of it. As a whole my argument is not about the problem you're describing. My argument is saying that the problem is a sham. An illusion. The problem doesn't even exist. Read my whole write up.

Yeah I stopped reading since it's based on a false predicate. I've read the rest and it's still undermined by the same unprovable assumption.

Perhaps consciousness is binary and the rock is just as conscious as a human. We don't know. But consciousness != intelligence, and it's reasonable to assume humans are more intelligent than rocks of course, but we can't say anything about each one's level of consciousness.

Re: Large language models lack deep insights or a theory of mind

#249

Post by mdp2021 »

educaysean wrote: post #110
mdp2021 wrote: post #109

> A chief goal of artificial intelligence [would be] to build machines that think like people "A chief goal of levers (cranes, etc.) engineering would be to build devices that lift like people"

Well we are the most intelligent species known to us as of now. Of course it would be considered the holy grail of simulated intelligence.

Reality instead is, that humans have a potential for Intelligent thought and action, and that they overwhelmingly avoid it with clamorous failures. To simulate that routine would be the opposite of the goal.

There is gain in implementing desirable qualities. Just that.

Re: Large language models lack deep insights or a theory of mind

#250

Post by FrustratedMonky »

krainboltgreene wrote: post #171
FrustratedMonky wrote: post #157

Earlier quoted context omitted.

Exactly. 5 years ago, we would have said what GPT-4 is doing now would be AGI. Now it is here and it's like "No, what we really meant is it has to be the next Einstein". People are forgetting how stupid people are. GPT is already better than average human. Most people can't do what we claim GPT must be capable of to qualify as AGI. The only logical conclusion is that many people are also not conscious and don't quali…

> Exactly. 5 years ago, we would have said what GPT-4 is doing now would be AGI. okay, and 500 years ago we would have said it was magic, that doesn't make it magic. people who don't understand the thing often are confused about the thing. as soon as you explain how the whole mechanism works it's obvious that it's not that thing. > People are forgetting how stupid people are. GPT is already better than average human.…

Idiocracy wasn't a documentary? Have you worked with people?

This sounds like someone sitting in their study, by the fire, smoking a pipe, adjusting their monocle and saying "hmm, Indubitably".

Think you are taking extreme opposite view, like humans have some holy essence that can't be understood or duplicated. Which is also not correct. Of course, technology that is not-understood can appear as magic. But nobody said anything of the sort.

The discussion was around AGI, will AI be as 'intelligent(loaded term)' as human. And I'm saying, a lot of people aren't doing anything that takes much 'intelligence' and current AI is already good enough to replace a large chunk of humans. Right now it is fragmented, there is AI that can beat GO, there is AI that can outfly an F-16 and beat human pilots, there is AI GPT/LLMs. It wont take much to start tying together technologies and have something that 'appears' human. And if it is indistinguishable, then how can we prove it isn't conscious, or opposite, prove that we are.

Post reply on HN