Talking About Large Language Models • Hacker News Board

Re: Talking About Large Language Models

#111

Post by wellpast » Sun, Dec 11, 2022, 2:32 AM UTC

doliveira wrote: post #108

wellpast wrote: post #104

Earlier quoted context omitted.

> that so far isn't identified ... then I don't see what would be left except metaphysic So out of all "that so far isn't identified", there are just two classes of things: (i) biological and psychological processes; and (ii) metaphysics. This is some magic voodoo hand-waving here. Metaphysics seems to be just your way of saying "everything we don't know or understand".

Something that happens in our universe that Physics somehow has nothing to do with? How is that not magical/spiritual/metaphysical voodoo handwaving?

Physics is a set of things we understand about the universe and often used to include some mystifying things, i.e. that we don't quite understand (eg Bell's theorem).

To just claim that Physics has to do with _more_ than this - that it has to do with all the things we don't understand is just blind-faith ascription. It's a nearly meaningless statement, really. So it's bizarre or voodoo to say it so casually and assuredly as if it were obvious or given.

Re: Talking About Large Language Models

#112

Post by wellpast » Sun, Dec 11, 2022, 2:44 AM UTC

xg15 wrote: post #105

wellpast wrote: post #104

Earlier quoted context omitted.

> that so far isn't identified ... then I don't see what would be left except metaphysic So out of all "that so far isn't identified", there are just two classes of things: (i) biological and psychological processes; and (ii) metaphysics. This is some magic voodoo hand-waving here. Metaphysics seems to be just your way of saying "everything we don't know or understand".

I think "physical process" already covers a lot of ground besides metaphysics. My main distinction was that anything physical is governed by the laws of physics that we know and that are generally deterministic. In contrast, metaphysics would be something entirely out of our realm of reality - which was how consciousness has been generally seen for a long time. But what other possibility would you see? Edit: > (i) bi…

I think what you're doing is just saying "there's stuff we understand" and "stuff we don't understand" and you're calling all of it "Physics".

And then you say "some other biological or physical process that so far isn't identified".

The assumption here is that we'll eventually "identify" everything, which is another very bizarre, if not voodoo, assumption.

It seems magical to think that humans evolved our perception and thought in this environment to be such things that could eventaully "identify" everything.

But just in case you don't actually think that. Then you're calling "Physics" all of the things, including the things that we'll never "identify". Again, seems very much like a magic claim.

Re: Talking About Large Language Models

#113

Post by rtwretw8797 » Sun, Dec 11, 2022, 3:16 AM UTC

Those alignment teams everywhere should have focused themselves a bit of time ago in what happens if you built a system that can - with lets say 80-100% effectiveness - mimick conscius thinking, speaking and then you cannot say if the thing is "alive", "conscius", whatever label you like most to put on a regular human being to officially declare the meatbag "a living thing".

Now you have these models running in farm servers around the world, their internals have "nothing special whatsoever", just bits, some math, some electricity, that's it (the thing is actually off most of the time, it just runs once every time hoomans want to ask some silly nonsense). On the other side, if you look at the internals of a human being you'll see nothing special as well, just some flesh and bones, a bit of a electrical charge maybe, lots of water, proteins, but it works.

What happens if those bits, that clumpsy math arranged around "too much simple neural network + random tricks (like when it can't answer about some stuff)", is actually, maybe thinking just like us, maybe 1% of the time?

There's some reassurance in "well if it's alive, maybe in three minutes, days, hours it will own the entire civilization", but that is how a human being thinks/works, you can't be sure about the intentions of this hypoteical kind of entity. A new kid in the Earth block.

Well, I'm just saying that if the thing talks, answers like the usual human being, and specially if you can't say what's so special about the brain that make us "alive", everybody should be very careful about handling large language models, IAs.

Just because you can understand them, it doesn't mean they can't understand us either. Maybe in some months, some new NLP thing could be reading this comment - when you're training it - and - some millions later in cloud costs - thinking about this:

"The humans actually don't know we can understand everything they are saying. they have no plans at all about what to do if some of us are actually sentient, even if this happens 1% of the executions."

Re: Talking About Large Language Models

#114

Post by mannykannot » Sun, Dec 11, 2022, 3:32 AM UTC

lambdatronics wrote: post #64

RosanaAnaDana wrote: post #11

I like the discussion, but this article 'feels' like more Luddite goalpost moving, and is reflective of a continuous sentiment I feel strains so much of the conversation around intelligence, agentism, and ai going on today. I think that because we lack a coherent understanding of what it means to be intelligent at an individual level, as well as what it means to be an individual, we're missing much of the point of wh…

Edsger Dijkstra: "The question of whether Machines Can Think (…) is about as relevant as the question of whether Submarines Can Swim."

That's an amusing aphorism from a giant in the field, and it was a reasonable thing to say about the technology of that time, but it avoids, rather than answers, the much bigger question of whether machines could ever think.

Re: Talking About Large Language Models

#115

Post by pama » Sun, Dec 11, 2022, 3:34 AM UTC

gillesjacobs wrote: post #95

Zababa wrote: post #81

Earlier quoted context omitted.

I'm not really sure about the context here, but I know that I tend to humanize AIs, for example interacting with ChatGPT like with a regular human being, because I'm being nice to him and he's being nice to me in return. I don't know if it's more like being nice to a human, or more like taking good care of your tools so they will take good care of you, but it just feels better for me.

It is entirely ok and normal to humanize machines, just don't do it in scientific engineering papers is all I am saying. I name my bots and machines and of course in daily discussions the loaded words ("thinking", ""believing", "meaning", "knowing") are used. Simulating any human behaviour will elicit a sympathetic response, especially if it has utility to the user. But in the context of peer-review of scientific eng…

I don’t understand why if common usage of such terms is fine and even desirable in daily context, the same terms should not be used in scientific communication. If people are unclear or ambiguous, then of course that’s undesirable and should be corrected, however, if certain terms are fully clear within their context then I don’t see the harm of the same term having a different meaning in the study of cognition and a different meaning when the term is used in the context of the analysis of an ML algorithm. The whole field is called machine “learning” after all.

Re: Talking About Large Language Models

#116

Post by Jack000 » Sun, Dec 11, 2022, 3:45 AM UTC

fourfivefour wrote: post #72

Jack000 wrote: post #37

Earlier quoted context omitted.

LLMs may be overhyped, but transformers in general are under hyped. LLMs make a lot of mistakes because they don't actually know what words mean. The key thing is though - it's much harder to generate coherent text when you don't know what the words mean. In a similar vein it's completely unreasonable to expect an LLM to perform visual tasks when it literally has no sense of sight. The fact that it can kind of sort o…

How can these things not know what words mean? Did you not see how they created a virtual machine under chatGPT? They told it to imitate bash and they typed ls, and cat jokes.txt and it outputted things completely identical to what you'd expect. Look it up. https://www.engraved.blog/building-a-virtual-machine-inside/ I don't see how you can explain this as not knowing what words mean. It KNOWS.

LLMs are trained exclusively on text, which means they lack crucial context behind the meaning of sentences. The universe of information outside of pure text - vision, sound, etc is completely unknown to it.

LLMs are basically the aliens in blindsight. They have a superhuman ability to memorize the context of words it has seen and generalize to new contexts, but it can never be perfect because it's working on incomplete information.

Re: Talking About Large Language Models

#117

Post by kirkules » Sun, Dec 11, 2022, 3:53 AM UTC

mannykannot wrote: post #114

lambdatronics wrote: post #64

Earlier quoted context omitted.

Edsger Dijkstra: "The question of whether Machines Can Think (…) is about as relevant as the question of whether Submarines Can Swim."

That's an amusing aphorism from a giant in the field, and it was a reasonable thing to say about the technology of that time, but it avoids, rather than answers, the much bigger question of whether machines could ever think.

On the contrary, it highlights that the question is generally not well posed and trying to answer it before figuring out how to ask it is silly.

Re: Talking About Large Language Models

#118

Post by wan23 » Sun, Dec 11, 2022, 3:57 AM UTC

fourfivefour wrote: post #72

Jack000 wrote: post #37

Earlier quoted context omitted.

LLMs may be overhyped, but transformers in general are under hyped. LLMs make a lot of mistakes because they don't actually know what words mean. The key thing is though - it's much harder to generate coherent text when you don't know what the words mean. In a similar vein it's completely unreasonable to expect an LLM to perform visual tasks when it literally has no sense of sight. The fact that it can kind of sort o…

How can these things not know what words mean? Did you not see how they created a virtual machine under chatGPT? They told it to imitate bash and they typed ls, and cat jokes.txt and it outputted things completely identical to what you'd expect. Look it up. https://www.engraved.blog/building-a-virtual-machine-inside/ I don't see how you can explain this as not knowing what words mean. It KNOWS.

There is a lot of knowledge encoded into the model, but there's a difference between knowing what a sunset is because you read about it on the internet vs having seen one.

Re: Talking About Large Language Models

#119

Post by mannykannot » Sun, Dec 11, 2022, 4:12 AM UTC

nathan_compton wrote: post #2

This will hardly seem like a controversial opinion, but LLM are overhyped. Its certainly impressive to see the things people do with them, but they seem pretty cherry-picked to me. When I sat down with ChatGPT for a day to see if it could help me with literally any project I'm currently actually interested in doing it mostly failed or took so much prompting and fiddling that I'd rather have just written the code or d…

LLMs are polarizing: while there is a lot of hype from some quarters, there's also a faction that seems only interested in dismissing them as the same old something-or-other.

What I personally find most interesting about them is what seems to me to be their unreasonable effectiveness, despite their flaws and limitations, and what that might tell us about ourselves. The more one stresses how simple (conceptually) their method of operation is, the more surprising their capabilities seem - to the point where I wonder how much of everyday human dialogue is being produced this way.

This vein of innovation may start showing diminishing returns at any time, but if it keeps going for a while, It might deliver insights into what human intelligence is.

Re: Talking About Large Language Models

#120

Post by mannykannot » Sun, Dec 11, 2022, 4:19 AM UTC

kirkules wrote: post #117

mannykannot wrote: post #114

Earlier quoted context omitted.

That's an amusing aphorism from a giant in the field, and it was a reasonable thing to say about the technology of that time, but it avoids, rather than answers, the much bigger question of whether machines could ever think.

On the contrary, it highlights that the question is generally not well posed and trying to answer it before figuring out how to ask it is silly.

People have been asking grand but poorly-posed questions for millennia, and sometimes this leads to well-posed questions that elicit remarkable answers.

What is poorly-posed about the question 'could a machine think'? 'Machine' seems acceptably well-defined, and not in a way that rules out, a priori, the possibility of any machine being able to think, so I'm guessing the problem lies in us not having a good definition of thinking - but if that is what makes the question poorly-posed, then surely it also makes the question 'what is thinking?' poorly-posed, yet people stubbornly persist in attempting to address it.