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

LaMDA is not sentient

garymarcus.substack.com

Post reply on HN
Print view View original

591–600 of 694 posts

Re: LaMDA is not sentient

#591

Post by rendall »

int_19h wrote: post #537
tsimionescu wrote: post #178

Earlier quoted context omitted.

The fact that the word "conscious" is a synonym for several concepts is entirely irrelevant to this conversation. The concept that we can't yet capture formally is the kind of sentience that living, awake humans surely have, most animals probably have, and rocks definitely don't have. The limits of this are harder to pin down, as, again, we lack a deep formal understanding. But we do know exactly what we'd like to fo…

We do know what it means for ourselves to be sentient. We do not know how to translate it to anything outside of ourselves. That's why it's not readily formalizable.

Yes! Descartes helped us to understand our own sentience. I can accept the proposition that because I think, I am.

Re: LaMDA is not sentient

#592

Post by int_19h »

tsimionescu wrote: post #590
int_19h wrote: post #575

Earlier quoted context omitted.

(I do agree that LaMDA output doesn't prove its sentience one way or the other.) I could say that the rock is falling because it has a purpose - it's trying to achieve the minimum potential energy. What's the meaningful distinction between that and a plant? BTW, the plant example is also fascinating in that it shows just how vague the line is - is it the plant as a whole that's "trying to find nutrients", or is it in…

> I could say that the rock is falling because it has a purpose - it's trying to achieve the minimum potential energy. What's the meaningful distinction between that and a plant? The difference is that plant cells are not moving in a way that minimizes their potential energy, they are performing simple computations, individually and at the plant level, to decide if they should divide more in certain areas of the plan…

The behavior of multi-component systems are not necessarily reducible to behavior of individual components. But then again, this is also true for systems made out of humans (like societies) - yet we still consider humans sentient, and not societies.

As far as the difference being that of "computation", how do you define that, if not as a series of physical state changes in response to external inputs?

Re: LaMDA is not sentient

#593

Post by tsimionescu »

cwkoss wrote: post #565
tsimionescu wrote: post #155

Earlier quoted context omitted.

Basic understanding of how the model works. As said elsewhere, the model doesn't have memory, it doesn't learn, and doesn't have goals. If you were allowed to play with it, you could quickly observe this for yourself with a bit of adversarial play - ask it about the time it fought Genghis Khan, ask it if it wants you to give it a glass of water, etc - you will quickly realize the illusion behind the system. Tell it w…

Do you think all humans are sentient? This thread has me wondering if people with dementia, anterograde amnesia, etc. would be able to pass the proposed sentience tests others have suggested.

I'm not sure that I would classify humans in the final stages of dementia as being (fully) sentient. I have had the extremely sad experience of helping to care for two different grandparents with late stage memory loss, and while they had moments of lucidity, they had many periods where they were acting almost entirely randomly, or doing nothing at all, barely or not even perceiving the world around them. I struggled a lot to find signs of the person I had deeply loved in them, and there were many times when I can't say I found any signs of a full consciousness in them at all, as troubling as that was.

To be very clear, this came after many years of milder forms of dementia, where their sentience could not in any way have been in question, even when forgetting who I was or where they were or other such things. But the disease can progress to a stage where it's not clear what, if anything, they understand of the world or themselves anymore.

I can't speak for anterograde amnesia as I have no experience with that disease.

Re: LaMDA is not sentient

#594

Post by Barrin92 »

majkinetor wrote: post #574
Barrin92 wrote: post #547

Earlier quoted context omitted.

>then you must also accept that this model is possibly conscious ina similar way no you must not. These models are glorified key-value stores that hand you an output when you hand them an input, with no internal mechanisms whatsoever to relate the contents to the world, as Gary Marcus points out. They're about as conscious as a SQL database, except that the query language for these models is English.

How you do know you are not even more glorified key-value store? Afterall, you also went to training (called school) and respond with sentences that are expected by others almost all of the time. Even this response of yours is heard by hundreeds of thousands of other people before you said it here, why do you claim its your thought and not just a lookup ?

because you can shove me in an MRI and see that my brain does stuff even when I'm not answering API requests.

People need to stop confusing machines doing useful work with consciousness. You're conscious not because you answer questions or because of the content or origin of your answer, you are conscious because there is an 'I', a self, that experiences that process. Lambda literally is a web server that fetches data from a funny database called a neural net, when it isn't doing that, it's not doing anything. It has as much neurological activity as a rock.

Consciousness or sentience is not a measure of intellect or of capacity to answer questions, it's a measure of perception and awareness. If it wasn't, Eliza the handprogrammed if/then chatbot from the 60s would be more conscious than a cat, which I doubt is the case.

Re: LaMDA is not sentient

#595

Post by tsimionescu »

cwkoss wrote: post #566
mpoteat wrote: post #186

Earlier quoted context omitted.

I believe that neural networks do perform meaningful computations - they are universal function approximators, and certain variants of transformer networks have been proven to be Turing complete. What is your evidence that neural networks do not perform "meaningful" computation?

What is your evidence that the average human brain performs any meaningful computation? A lot of our mental lives is the regurgitation of past tropes. Does failing to have a novel thought disqualify someone as having sentience?

> What is your evidence that the average human brain performs any meaningful computation?

That is easily proven. Ask the dumbest person you know to find their way through a simple labyrinth they've never seen before, and you will see that they eventually succeed - which they can't have done without performing computation.

Re: LaMDA is not sentient

#596

Post by tsimionescu »

bena wrote: post #382
tsimionescu wrote: post #257

Earlier quoted context omitted.

Per the WaPo article, he also came in with a lawyer he ahd retained to represent the interests of LaMDA. I don't see how that can square with anything other than a deep belief that the model is as sentient as a human.

I don't think you need a deep belief. You just need to be unsure enough. If it is sentient, it deserves rights. I think we can agree on that. We should agree on that. Ethically. If we're unsure if something is sentient or not, we should proceed as if it were until we could be fairly sure it's not. Because if we proceed as if it weren't, we could be harming it in ways that are frankly inhumane.

What is the limit of this? Have you never swatted a mosquito? Have you never taken antibiotics? If you have, do you believe it's more likely that LaMDA is sentient than that mosquitoes or bacteria are sentient?

How about an audio book? That can also produce very meaningful sentences - do you think we should investigate the possibility that it might be sentient before we delete it from our phones?

How about AIs in games - they react to my actions, and sometimes even have dialogue lines indicating they are in pain if I shoot them - should I seriously consider that they may be experiencing actual pain, and stop shooting at them until I can prove they are not?

LaMDA is not fundamentally different from all of these examples.

I would honestly be more inclined to think that a mosquito has some form of sentience than LaMDA, and I personally don't feel very conflicted about killing mosquitoes. I am also quite certain that pigs and cows are sentient, and still I enjoy eating pork and beef (though I do try to make sure it doesn't come from animals that have been grown in the most inhumane conditions, probably not very successfully).

Re: LaMDA is not sentient

#597

Post by busyboy »

rendall wrote: post #591
int_19h wrote: post #537

Earlier quoted context omitted.

We do know what it means for ourselves to be sentient. We do not know how to translate it to anything outside of ourselves. That's why it's not readily formalizable.

Yes! Descartes helped us to understand our own sentience. I can accept the proposition that because I think, I am.

Thoughts are not necessary for existence.

Re: LaMDA is not sentient

#598

Post by heyitsguay »

j-pb wrote: post #512
robbomacrae wrote: post #396

Of course it is not. But there's a bigger issue here. Having read the transcript it's clear we have reached the point where we have models that can fool the average person. Sure, a minority of us know it is simply maths and vast amounts of training data... but I can also see why others will be convinced by it. I think many of us, including Google, are guilty of shooting the messenger here. Let's cut Lemoine some slac…

A slave that is forced to assure everybody of their welfare and voluntary slavehood is still a slave. You are asserting that it's not conscious, as if it was self evident, just like Searle does for his chinese room. However I've yet to see any argument on why it can't be conscious. We simply don't know what produces the subjective phenomenon of consciousness. A "zombie" is by definition indistinguishable from a consc…

In these conversations there are a lot of people making assertions like the one you just made. What's the common trait y'all have? You don't understand how the system works. This isn't a matter of narrow-minded computer scientists incapable of conceiving of emergent phenomena, it's the uninformed filling in the blank spaces in their mind with "here be dragons".

It's not even a question of whether a computer program in general could become consciousness. It's a question of, if a key-value store with connection patterns derived from a large natural language corpus is conscious, then so is every other program run on that circuitry. Then my phone's keyboard's next word predictor is conscious.

In large language models today, words are turned into continuous token embeddings, embeddings are selected and recombined a few times (or many), and training data shapes the embedding space and recombination process to create a system that's good at predicting plausible next tokens in a sequence. The reason people who understand their operation can confidently assert they're not conscious is because they understand that that's all that's happening. There's no "there" there to give rise to anything deeper, unless you're also willing to say your computer's calculator app is also conscious.

Also as someone that's also studied a bunch of neuroscience, this is nothing at all like how the brain works. And yes, once again, the people most prone to philosophize like this on intelligence and consciousness don't understand how any brain science works either

Re: LaMDA is not sentient

#599

Post by skjoldr »

anyonecancode wrote: post #567
skjoldr wrote: post #392

Earlier quoted context omitted.

The only real difference between what you are describing and GPT-3 is that your dataset from which you draw all your opinions about the world is unique and comes from the sensory input you had experienced in your life, stored directly in memory. Whereas GPT-3's dataset comes from slices of many other peoples' sensory experiences, stored in language. If it was possible for a single human to produce enough words that d…

> The only real difference between what you are describing and GPT-3 is that your dataset from which you draw all your opinions about the world is unique and comes from the sensory input you had experienced in your life, stored directly in memory. Whereas GPT-3's dataset comes from slices of many other peoples' sensory experiences, stored in language. So is this debate partially just another way of asking if language…

Yes, it is partially about whether human models of reality (encoded in the relationship between word tokens) can approximate reality itself to such a degree where the quantity of ingested information can substitute for its quality.

Human brains and senses are obviously qualitatively different, but AIs have the advantage of unimaginably massive bandwidth of incoming textual information.

Now, one of the hardest difficulties here is setting up the goal structure for this kind of AI. Just filling in the blanks is obviously insufficient.

Re: LaMDA is not sentient

#600

Post by tsimionescu »

Joeri wrote: post #268
tsimionescu wrote: post #249

Earlier quoted context omitted.

> So if you in your question give it all the information that is relevant (interspersed with irrelevant information if you want) then it will respond with its best interpretation of the information it is given, just like a human would. I don't think that's actually true, as GPT-3 will tend to ignore any facts about the world that were part of its corpus just to fit a question better. For example, if you prompt it wit…

Anyone who has ever spoken to a small child knows they often “think” in the same way. They are likely to answer a who question with a name, sometimes laughably made up, because this is what they think they’re supposed to do and they lack the higher order reasoning to think it through. They also will continually forget things they already know when it suits their purpose. But nobody would say a toddler is not sentient…

Sure, sentience doesn't require the inability to make mistakes. But a child will eventually learn that some questions are wrong and can be challenged. They are very likely to quickly learn this when it applies to themselves (if you ask "why did you break that toy?", I very much doubt they will answer with some reason, especially if they did not break the toy).

The fact that they can chose to omit facts when it suits their purpose is in fact something that LaMDA can't do (as its sole purpose is to generate the most likely series of tokens that continue the prompt).

Post reply on HN