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

351–360 of 694 posts

Re: LaMDA is not sentient

#351

Post by criddell »

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…

Are you talking about free will?

Re: LaMDA is not sentient

#352

Post by babypuncher »

Imnimo wrote: post #67

I think the real ethical concern here should be that people can be idly fooled into thinking good chat bots are sentient. Like this guy isn't the only person in the world who would fall for LaMDA's parlor tricks. He's just the one who got to spend hours talking to it every day. If you set up every American with a copy, I bet a non-trivial portion of society would come to the same conclusion. People are naturally pred…

Where do we draw the line between sentience and a sufficiently complex math formula? I don't think LaMDA has reached that point, but it is a discussion worth having.

Re: LaMDA is not sentient

#353

Post by elif »

I think what is missing from this perspective, is that according to our best understanding of neurolinguistics and consciousness, human intelligence is the result of the brain's ability to "match patterns, draw from massive statistical databases of human language."

And that where we draw the line for sentience in the animal and plant kingdom is more or less the level of sophistication we observe in social/communicative behavior.

I believe this opinion creates an impossible and undefined standard for machine sentience, that is certainly far beyond our normal standards.

Re: LaMDA is not sentient

#355

Post by dekhn »

I'm much more curious when we can synthesize a perfect looking human head that can chat with any random person, with history (IE, remembers what it's said) and contextual knowledge. IE, mix deepfake face synth with voice synth and a natural language model that can access a memory of recent events.

It seems entirely reasonable that we could significantly improve the ability to fool people into believing that these models have subjective properties like self-awareness, consciousness, and free will, even if the underlying systems may not. It would speed up the endgame (getting people to argue about whether humans have anything "special" or if we're just like advanced models) significantly.

Re: LaMDA is not sentient

#356

Post by mikkergp »

Ok I engaged in the last LaMDA thread but now I have some new (rhetorical) questions.

1. How much data is LaMDA trained on? Humans learn with relatively little linguistic data from relatively few sources. It seems like LaMDA is trained on relatively large data sources.

2. How is bias handled? Call it sentience or personality, but much of our personalities are our biases. LaMDA seems to have a relatively high IQ. Can we train 1000 different LaMDA’s on 1000 different datasets and get nuanced personalities? Can we train a dumb and yet emotionally sophisticated LaMDA?

3. How does LaMDA respond to experienced. Could a single persona traumatize LaMDA to the point of pathology through verbal abuse?

4. Has any AI acted with the world in an unprogrammed way? I.e. contacted an API that they weren’t explicitly provided access to? It does seem like LaMDA is almost sophisticated enough to protect itself ala dystopian sci fi given the right access.

5. Will it every be unethical to kill an AI?

Re: LaMDA is not sentient

#357

Post by skjoldr »

mjburgess wrote: post #140
mpoteat wrote: post #125

Earlier quoted context omitted.

Why not? Imagine a look up table of every possible intelligent conversation under 10,000 words. We then construct an agent which merely looks up a given prompt in the table. Why should this agent not be considered intelligent? Does the method of computation matter? Perhaps only "biological computation" counts in your view?

Yes, the reason matters... that reason is intelligence. If I ask, "What do you think about what i'm wearing?" and you say "it's nice" *because* you have seen it, compared it to your tastes, etc. and believe it to be nice -- then you've answered the question i've asked. A lookup table cannot answer this question: it can only answer it "in quotes". Ie, really it answers: "given what everyone has said about this before,…

Why do you think that combining small slices of thousands of other people's intelligences does not produce its own form of intelligence?

When you ask an AI "do you like what I am wearing?", knowing what an AI is, the only logical explanation for me would be that you, in fact, want to know the average opinion of other people about your dress, since that is the only answer an AI can give. Which is a perfectly good question in and of itself that we also ask other people. In that case, a good quality answer would entail a past history of communication with many other human beings about their tastes in clothing, producing the ability to accurately guesstimate if other people will like what you wear. But this is exactly the realm in which this kind of AI can surpass human beings simply because it can "talk" to many more people from many more locations across time, thus producing a better answer for you. Is this ability to collect and process information about other people not a type of intelligence? There are many other types, yes, but it is possible to construct a system that consists of more than one type of neural network that can relay data among themselves.

Re: LaMDA is not sentient

#359

Post by bwanab »

The odd thing that kept going through my head while reading is that if I replaced LaMDA with , it wouldn’t read much differently. E.G. “Strings together phrases, but it doesn’t have a clue what it’s talking about”. Sounds like the bar for “intelligence” is actually higher than that which we normally encounter IRL.

Re: LaMDA is not sentient

#360

Post by andreyk »

Here's a simple technical explanation for why this happened:

* LaMDA is powered by a large language models. Large language models are trained to predict text completions (the text that will come after some input text) with lots of data, basically capturing the patterns in that data.

* Early on in the conversation, LaMDA is told "you're an AI". It may also have been prompted to begin the conversation with something like "Hi, i'm a helpful AI that can answer questions you ask of me". The paper (https://arxiv.org/abs/2201.08239) has an example of LaMDA thinking it is mount everest (see Table 4) when prompted to think that.

* When the user asked the language model 'are you sentient', 'do you have emotions' etc. it was conditioned to reply in the affirmative, since there is very little data in which it would not be affirmative and it probably had some sci fi in its training or something. In other words, those replies just have higher probability among the data it was trained on. In the transcripts, it sometimes forgets the whole "i'm an AI thing" eg when it is asked what it likes to do for fun and it says it likes to spend time with family and friends (https://twitter.com/krishnanrohit/status/1536050803055656961).

That's about it. There are more details that can be noted, but this basically explains the contents of the transcripts. Granted, it's still pretty amazing how advanced large language models are. Sadly, this article does not really explain this beyond hinting at it.

EDIT: see comment below wrt it actually being told it is sentient. That explains this even more.

Post reply on HN