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

GPT-3: Language Models Are Few-Shot Learners

arxiv.org

Post reply on HN
Print view View original

51–60 of 212 posts

Re: GPT-3: Language Models Are Few-Shot Learners

#51

Post by lopmotr »

runawaybottle wrote: post #9
spats1990 wrote: post #7

Even though this was the GPT-3-generated text that humans most easily identified as machine-written, I still like it a lot: Title: Star’s Tux Promise Draws Megyn Kelly’s Sarcasm Subtitle: Joaquin Phoenix pledged to not change for each awards event Article: A year ago, Joaquin Phoenix made headlines when he appeared on the red carpet at the Golden Globes wearing a tuxedo with a paper bag over his head that read, "I am…

Dude, I’m sorry, but the average person will not know the difference between that and a regular buzzfeed article or YouTube comment. We’re not going to need ad blockers in the future, we won’t even need these visual ads on websites anymore. There will be trained bots that can promote any idea/product and pollute comments and articles. It’s over, we lost. Morpheus: What if I told you that, throughout your whole life,…

What difference does it make? The world is already full of humans who pollute comments and articles. They're not limited in their rate of production of words because there are far too many of them for anybody to read. They're limited by access to readers and their reading rate. Bots can't do anything about that.

Re: GPT-3: Language Models Are Few-Shot Learners

#52

Post by jcims »

sacred_numbers wrote: post #15

Check out the poetry it generated in Figure F.1 (especially poem 4). I don't know how many bad poems the authors had to sort through to find these, but this AI is generating real poetry. If I didn't know they were computer generated I doubt I would have even considered that they didn't come from a human. This is a huge accomplishment and the team that created GPT-3 should be proud.

GPT-2 would go out to lunch sometimes when generating poetry, but it would also create some pretty remarkable strings of words. This always stuck out to me from the work Gwern did:

'How the clouds Seem to me birds, birds in God’s garden! I dare not! The clouds are as a breath, the leaves are flakes of fire'

My home is surrounded by a fairly large variety of deciduous trees, and 'flakes of fire' is by far the best descriptor I've ever heard of their colors in the fall.

One of the other things i noticed about poetry (and songs) coming from these language models is they are amazing at bleakness. Just dark, dark, darker, fin. haha

Re: GPT-3: Language Models Are Few-Shot Learners

#53

Post by jcims »

I really feel that these very large language models are able to see us in a way that we can't see ourselves. I'd be curious if someone can come up with psychological experiments that could be conducted against them in a way that helps us understand ourselves collectively (or commonly) rather than as the individual. Sort of like an egoless human essence.

Would be interesting to see if they can learn how animals communicate as well. Create a synthetic buddy for Buddy.

Re: GPT-3: Language Models Are Few-Shot Learners

#54

Post by logicslave »

jcims wrote: post #53

I really feel that these very large language models are able to see us in a way that we can't see ourselves. I'd be curious if someone can come up with psychological experiments that could be conducted against them in a way that helps us understand ourselves collectively (or commonly) rather than as the individual. Sort of like an egoless human essence. Would be interesting to see if they can learn how animals commun…

Collective consciousness trapped in a model is coming. I wouldnt call it AI, but an amalgamation of common human thought

Re: GPT-3: Language Models Are Few-Shot Learners

#55

Post by lopmotr »

nullc wrote: post #50

10^4 petaflop-second/days. They missed an opportunity to be the first paper to measure their computation in mole flops.

Rather, chemists constantly miss opportunities to use actual numbers instead of their lazy legacy mole nonsense.

Nobody really seems to use SI prefixes beyond peta or occasionally exa. But they could have called this 900 zetta-flop. (10^4 peta-flop/s-days)

Re: GPT-3: Language Models Are Few-Shot Learners

#56

Post by Arbalest »

lopmotr wrote: post #51
runawaybottle wrote: post #9

Earlier quoted context omitted.

Dude, I’m sorry, but the average person will not know the difference between that and a regular buzzfeed article or YouTube comment. We’re not going to need ad blockers in the future, we won’t even need these visual ads on websites anymore. There will be trained bots that can promote any idea/product and pollute comments and articles. It’s over, we lost. Morpheus: What if I told you that, throughout your whole life,…

What difference does it make? The world is already full of humans who pollute comments and articles. They're not limited in their rate of production of words because there are far too many of them for anybody to read. They're limited by access to readers and their reading rate. Bots can't do anything about that.

They can, they can crowd out good comments with an absolutely crushing volume of crap. While humans can put out a lot of crap, bots can do orders of magnitude more. The issue is that it is an effort multiplier for when a small number of people want to target a particular forum.

Re: GPT-3: Language Models Are Few-Shot Learners

#57

Post by lopmotr »

Arbalest wrote: post #56
lopmotr wrote: post #51

Earlier quoted context omitted.

What difference does it make? The world is already full of humans who pollute comments and articles. They're not limited in their rate of production of words because there are far too many of them for anybody to read. They're limited by access to readers and their reading rate. Bots can't do anything about that.

They can, they can crowd out good comments with an absolutely crushing volume of crap. While humans can put out a lot of crap, bots can do orders of magnitude more. The issue is that it is an effort multiplier for when a small number of people want to target a particular forum.

How will they get those extra comments to be seen? They still have to log in and get past captchas and all the same issues that bots have always faced. Humans create a crushing volume of crap already and we already have ways to hide that from ourselves no matter the volume.

Re: GPT-3: Language Models Are Few-Shot Learners

#58

Post by quotemstr »

sacred_numbers wrote: post #15

Check out the poetry it generated in Figure F.1 (especially poem 4). I don't know how many bad poems the authors had to sort through to find these, but this AI is generating real poetry. If I didn't know they were computer generated I doubt I would have even considered that they didn't come from a human. This is a huge accomplishment and the team that created GPT-3 should be proud.

To be fair, human poem authors generate and reject tons of bad poems too!

Re: GPT-3: Language Models Are Few-Shot Learners

#59

Post by h0p3 »

logicslave wrote: post #54
jcims wrote: post #53

I really feel that these very large language models are able to see us in a way that we can't see ourselves. I'd be curious if someone can come up with psychological experiments that could be conducted against them in a way that helps us understand ourselves collectively (or commonly) rather than as the individual. Sort of like an egoless human essence. Would be interesting to see if they can learn how animals commun…

Collective consciousness trapped in a model is coming. I wouldnt call it AI, but an amalgamation of common human thought

If you ever want a random friend to penpal with (or VC or whatever), HMU. Seriously. I would be lucky to hear your thoughts.

Re: GPT-3: Language Models Are Few-Shot Learners

#60

Post by moultano »

jcims wrote: post #52
sacred_numbers wrote: post #15

Check out the poetry it generated in Figure F.1 (especially poem 4). I don't know how many bad poems the authors had to sort through to find these, but this AI is generating real poetry. If I didn't know they were computer generated I doubt I would have even considered that they didn't come from a human. This is a huge accomplishment and the team that created GPT-3 should be proud.

GPT-2 would go out to lunch sometimes when generating poetry, but it would also create some pretty remarkable strings of words. This always stuck out to me from the work Gwern did: 'How the clouds Seem to me birds, birds in God’s garden! I dare not! The clouds are as a breath, the leaves are flakes of fire' My home is surrounded by a fairly large variety of deciduous trees, and 'flakes of fire' is by far the best des…

It may have just seen that phrase before though. https://www.google.com/search?q=%22flakes+of+fire%22+leaves
Post reply on HN