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

311–320 of 694 posts

Re: LaMDA is not sentient

#311

Post by yencabulator »

snowwrestler wrote: post #66
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…

I have to wonder if he is a parent; I have a child around that age and there is zero chance they would sit around cheerfully answering questions from computer science researchers until they were done.

Some corvids are considered to be of equal intelligence as a 5-7 year old human child. That doesn't make them not-birds, and the human child isn't going to flap its wings and fly away.

Re: LaMDA is not sentient

#312

Post by blooalien »

jmmcd wrote: post #299
morelisp wrote: post #277

Earlier quoted context omitted.

For LLMs it's not really that interesting. There's no embodiment at all beyond text, there's no sense of "time" or "speed". Either it's running full-tilt generating a response, or it's off.

As I said, I think it's interesting if it uses the downtime for something else. (Like a chess player thinking on the opponent's click.) Talking to itself, speculating about continuations, etc - but not actually outputting them (else it's just a longer response). Instead, storing some parts in the buffer. Perhaps bifurcating. Even better, using some down time to summarise recent material and store the summary in the b…

Perhaps even occasionally, without being prompted by further input, continuing a prior statement with a further "followup thought" when some specific threshold of "speculation" is reached?

Re: LaMDA is not sentient

#313

Post by tcmart14 »

cs137 wrote: post #199
rlue wrote: post #129

Earlier quoted context omitted.

Everyone is giving Lemoine far too much credit. He's not an idiot; he's a charlatan. He's acting in bad faith and it should be extremely obvious to anyone who's tried to have a conversation with a chat bot before. Picture yourself sitting at a desk, typing messages into a program you wrote. You hit enter, you get a response. Then what? Silence. It waits idly for more input from you, the user. For all its talk of lone…

I really don't think this guy intended to run a hoax or generate publicity, given that he's made himself look foolish. I think he was actually seduced by the idea of a sentient computer. It's not that far a leap to go from the stance of well-respected people inside Google who believe we'll see artificial sentience within 30 years to believing that we've achieved it now with something like LaMDA. We're all suspectible…

Yea, it could be a case of the Eliza Effect.

Re: LaMDA is not sentient

#314

Post by morelisp »

jmmcd wrote: post #299
morelisp wrote: post #277

Earlier quoted context omitted.

For LLMs it's not really that interesting. There's no embodiment at all beyond text, there's no sense of "time" or "speed". Either it's running full-tilt generating a response, or it's off.

As I said, I think it's interesting if it uses the downtime for something else. (Like a chess player thinking on the opponent's click.) Talking to itself, speculating about continuations, etc - but not actually outputting them (else it's just a longer response). Instead, storing some parts in the buffer. Perhaps bifurcating. Even better, using some down time to summarise recent material and store the summary in the b…

But then it's not a regular LLM anymore and we don't really know how to build that.

The whole motivating aspect of these models is "attention is all you need" to reduce dependencies from recurrence, which would otherwise halt this obscene scaling in its tracks.

Re: LaMDA is not sentient

#315

Post by tcmart14 »

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…

There is a term that lines up with this called the Eliza Effect. And yea, it can definitely be an easy trap to fall into. Some people here are making suggestions he is mentally ill, he very well could be, but I dont think there is enough 'evidence' to show his claims are because of mental illness. Just as you mentioned, getting stuck talking to the thing and all it takes it to forget the implementation details. He did have an education in Computer Science, so I am pretty sure he is probably aware of the details, but yea, just getting sucked into the conversation can probably even cause someone who knows better to forget that part. Now as you said, what about when better chatbots become more common place. It is gonna be a mess since most people won't understand how these are implemented anyways.

Re: LaMDA is not sentient

#316

Post by renewiltord »

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…

I went to sleep last night, knowing that I would get a certain kind of misunderstanding of my comment, hence that last hash there. It's a pastebin URL (gives it a timestamp) https://pastebin.com/VT7L3TAb so you can compare it with the outcome condition and look at the timestamps.

That pastebin URL contains `65f7cae61536f9e4967ade278f77273daeff36a0f998853a7c9d7264bd4a9f4d`. A child comment[0] (since flagged) to this thread describes me one hour later:

> He's a narcissistic clown. He is fond of posting dehumanizing and insipid drivel, like the top-level comment here which denies the sentience of the wage slaves that serve him.

I pre-committed my answer to this reply before it was posted because I knew it would come: (verify using the hash and timestamp)

    echo "Ah, you're offended. But don't worry. I am not above being prompt-engineered too. That's the point" | sha256sum

    65f7cae61536f9e4967ade278f77273daeff36a0f998853a7c9d7264bd4a9f4d  -
In some sense this conversation chain (my original comment included) is a purely probabilistic output that one could mechanistically have created from sheer experience. So perhaps the transformers aren't as far from us as we think.

I actually knew that the other comments would also arrive and how I would respond, in general. It's rude to pre-commit to answers, though, so I could only safely do that to a guy who would be rude to me. The guy pulling my leg about Dr. Who did buck the chain (you bastard, ruining the point). I think we all have a degree of that. If you see "Google" and "service", you know the top comment "I won't ever use Google because they sunset Reader" or some such shit. These transformer AIs are rapidly approaching the level of HN-like social networks (like me and my comments) - and low-repetition interactions (like the above-mentioned pharmacists). The key thing is that they can't currently track specific state-alteration, i.e. they do not know this apple was moved from this side of the table to the other, but that's a sensory problem. We'll see.

0: https://news.ycombinator.com/item?id=31722264

Re: LaMDA is not sentient

#318

Post by stormsson »

jmmcd wrote: post #266
hackernews1134 wrote: post #230

Earlier quoted context omitted.

Why isn't it simply "asleep" when not questioned? In a "coma" perhaps. One made from our programming / "genetics"?!?! I don't think it is sentient without more evidence. But just wondering if not "thinking" when not questioned is a good reason it isn't?!?!

When it's not predicting a word, it's not doing anything. Literally GPU usage goes to zero. Still it's interesting! Obviously we have to program the system to stop after a sentence or two so the user can read it. What if we could program it to keep "talking to itself", eg speculating about possible continuations on both sides.

What about the time from the input received to the output? it's a total black box, for what we know it could "think" for 90% of that time and just use the last part of gpu power to produce the output (whatever "think" means in this context)

I don't think it should be reasonable to compare it with human reasoning where we think constantly and our brain operates even while we sleep: it is not a human being after all

Re: LaMDA is not sentient

#319

Post by null0ranje »

permo-w wrote: post #275
GalahiSimtam wrote: post #214

Earlier quoted context omitted.

we call it Voight-Kampff for short on a more serious note, there is a case for not being mean towards chatbots, same as not being violent towards patches of pixels in videogames, goes back to Immanuel Kant and animals, or so I've heard

yes there's the categorical imperative, but there's also a good chance that whatever you say to that chatbot is going to be used to train other AIs to be a dick to you

Or to train itself, to the worst possible outcomes: https://arstechnica.com/information-technology/2016/03/micro...
Post reply on HN