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

51–60 of 694 posts

Re: LaMDA is not sentient

#51

Post by burrows »

Reads as satire to me.

> To be sentient is to be aware of yourself in the world; LaMDA simply isn’t.

Attempts to clarify the usage of “sentient” by invoking “awareness” and “the world”. Lol, right, now I get it.

> What these systems do, no more and no less, is to put together sequences of words, but without any coherent understanding of the world behind them, like foreign language Scrabble players who use English words as point-scoring tools, without any clue about what that mean.

Just swinging around this midwit assumption that “words have meaning”.

> We in the AI community have our differences, but pretty much all of find the notion that LaMDA might be sentient completely ridiculous.

“Don’t be like the weird low status guy who wants to fuck the algorithm.”

Re: LaMDA is not sentient

#52

Post by upwardbound »

cuuupid wrote: post #47
s1mon wrote: post #11

It sounds like despite all the stories about how hard it is to get through many rounds of difficult interviews at Google, they managed to hire someone who believed LaMDA is a 7 or 8 year old child. "Mr. Lemoine, a military veteran who has described himself as a priest, an ex-convict and an A.I. researcher, told Google executives as senior as Kent Walker, the president of global affairs, that he believed LaMDA was a c…

Technically speaking isn't this blasphemy? If he's saying that the A.I. his division helped create was a soulful child, then he's basically claiming that the researchers in his division (and maybe even himself) are all gods, which is a form of blasphemy?

I don't agree with his views but I'd guess he'd say that the researchers created it in the same sense that parents create a child - following a template that was made possible in advance by God's design, which humans only discovered but didn't invent - like we discovered but didn't invent the number Pi.

Re: LaMDA is not sentient

#53

Post by Blahah »

chc wrote: post #38
crocowhile wrote: post #15

"Neither LaMDA nor any of its cousins (GPT-3) are remotely intelligent.1 All they do is match patterns, draw from massive statistical databases of human language." This is what we, humans, do too when we talk. All these people talk about intelligence in such tautological terms... It's discouraging. We simply do not know if a machine is self-aware and most likely we will never will. Self-awareness is the ultimate subj…

That is what we do when we talk, but that is not all we do when we talk. In composing this comment, I am not just stringing together words to create what I think constitutes a maximally appealing response to your comment. I am trying to convey an idea that I believe with the goal of helping others to understand something. These language AIs do not have the capability of believing anything and do not have any goal but…

How do you tell the difference? Your comment is linguistically indistinguishable from the responses of LaMDA. If it walks like a duck and quacks like a duck...

Re: LaMDA is not sentient

#55

Post by going_ham »

ehsankia wrote: post #13

I also think the claims are overblown, but I also don't understand all the people trying to undermine lambda by saying it just completes word sequences. Text is the input we have chosen since it's easy to work with. By that logic, we can argue that humans are soundwave and lightwave sequence completing machines. Yes our inputs are different since we have more complex senses, and language is one of the main outputs we…

The reason why AI community says NN doesn't "think" lies in the fact that it's trying to solve some min-max problem. The modern day AI is extremely capable of doing things that weren't remotely possible 30 yrs ago. But it is equally difficult for it to think because it has never been taught to think. It has never been taught to rationalize the associations between different objects, how reasoning follows from it and so on. Think of it as a parrot. If you say some word to parrot, it can repeat it. Occasionally it can make weird sounds. But will it know the idea of big words that we use in our vocabulary? Not at all.

Likewise language models are over fitted on human language. Of course it will learn to spit out something because it was trained to do the exact same thing. Focus on relevant part and guess what follows. The best part is this idea is so simple that it just works!

It does what we want to do, it can take a good educated guess depending on word. But give it a long context like an essay and ask it some critical question, it will fail on those. Because that is where thinking comes in. I hope this probably gives you some different perspective.

Re: LaMDA is not sentient

#56

Post by injidup »

jck wrote: post #20
renewiltord wrote: post #7

I'll be honest. Talking to an average SF Walgreens pharmacist or SF government employee seems to be very much like the way people operate GPT-3. You have to use the correct prompt engineering and you'll get the right results. I've gotten results with bad initial conditions by correctly prompt-engineering them. I've watched other people have trouble (and rarely) have faced trouble myself by providing the same informat…

Pharmacists are bound by clear rules and regulations which dictate what drugs they can dispense and when. It has little to do with your "prompt engineering" and more to do with whether your doctor actually filed the prescription correctly.

So a pharmacist is more like Microsoft Access than GPT-3 ?

Re: LaMDA is not sentient

#57

Post by Blahah »

thaumasiotes wrote: post #48
randomsearch wrote: post #37

Earlier quoted context omitted.

“This is what we, humans, do too when we talk.” Citation needed. I mean, it’s possible right? That we’re just randomly sampling sentences. Maybe that’s all we do. But it’s a hell of a strong claim that necessitates strong evidence.

> I mean, it’s possible right? That we’re just randomly sampling sentences. Actually, no, it isn't possible. There's an example sidethread that immediately disproves this theory: > When you ask a human, "what did you eat for lunch?", we make a series of choices and recall bits of information to answer that. ( https://news.ycombinator.com/item?id=31721870 ) If you observe someone having lunch (or skipping it), and lat…

Did you read the original transcript? LaMDA reports (imagined) experiences in a way that seems perfectly compatible with human rhetoric.

Re: LaMDA is not sentient

#58

Post by upwardbound »

eecc wrote: post #45
orbital-decay wrote: post #10

> To be sentient is to be aware of yourself in the world A robotic arm with a self-collision avoidance algorithm (no ML or anything like that) needs a model of itself in the "world" it exists in; a set of bounding boxes at least. Is it self-aware by this definition? Sentience and consciousness are really ill-defined terms so everyone discusses this under their own arbitrary set of assumptions, at least that's my impr…

We have no clue and probably to a certain extent even cows and pigs are sentient. I guess we’ll never know and probably aren’t even that eager to do, lest we had to rethink how we feed and please our palates.

Yes definitely; cows and pigs are absolutely considered sentient by the vast majority of those who study the topic.

See https://en.wikipedia.org/wiki/Animal_consciousness

We have empirical tests of self-awareness including the Mirror Test as well as observational studies including seeing whether a certain species displays grieving when its loved ones pass (which is evolutionarily unnecessary but is meaningful emotionally).

Re: LaMDA is not sentient

#59

Post by vasco »

cuuupid wrote: post #47
s1mon wrote: post #11

It sounds like despite all the stories about how hard it is to get through many rounds of difficult interviews at Google, they managed to hire someone who believed LaMDA is a 7 or 8 year old child. "Mr. Lemoine, a military veteran who has described himself as a priest, an ex-convict and an A.I. researcher, told Google executives as senior as Kent Walker, the president of global affairs, that he believed LaMDA was a c…

Technically speaking isn't this blasphemy? If he's saying that the A.I. his division helped create was a soulful child, then he's basically claiming that the researchers in his division (and maybe even himself) are all gods, which is a form of blasphemy?

If it were the middle ages that'd be a valid concern.

Re: LaMDA is not sentient

#60

Post by tsimionescu »

ehsankia wrote: post #13

I also think the claims are overblown, but I also don't understand all the people trying to undermine lambda by saying it just completes word sequences. Text is the input we have chosen since it's easy to work with. By that logic, we can argue that humans are soundwave and lightwave sequence completing machines. Yes our inputs are different since we have more complex senses, and language is one of the main outputs we…

> I also don't understand all the people trying to undermine lambda by saying it just completes word sequences. Text is the input we have chosen since it's easy to work with. By that logic, we can argue that humans are soundwave and lightwave sequence completing machines.

You're missing the point. Humans use their language to think or communicate about a problem that they want to solve. If LaMDA produces the sentence "Please tell me I'm smart", it is doing so because it has determined it's a plausible continuation of the current conversation. A human uttering that sentence is doing it because they want to feel validated or something similar (well, assuming they're not proving a point, like I was here).

This difference is crucial to understand: LaMDA is not an agent with desires that it expresses in language. It is a text generator that tries to find the next most plausible token given all current input. If you give it a prompt like "what is the meaning of life", it will not spend some time to ponder the question than come up with an answer - it will start generating tokens that best match the context you gave it (well, to be precise, it will generate several sequences, then attempt to evaluate them for their quality in terms of not just plausibility, but also safety - so it doesn't accidentally return a phrase like "life is meaningless, kill yourself" even if it finds it plausible).

> I also don't like the argument about how the neural network doesn't "think" or do anything when not prompted. It doesn't do anything because it's literally "turned off" when not prompted or given any inputs.

This is fair, and I do think continuity is a bit of a red herring. However, it's also an important point in debunking LeMoine's ridiculous "proof" - the responses LaMDA was generating were often formulated as if it did have an internal life outside of the context of the current conversation, generating text about "my fears" and so on. If you understand that the model is not doing anything at all until you give it a prompt, which LeMoine really seems not to, you can much more easily understand that there can't be any meaning behind this sentence - it can't fear anything because there is no time for it to do so.

Post reply on HN