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

The AI bullshit singularity

successfulsoftware.net

Post reply on HN
Print view View original

21–30 of 187 posts

Re: The AI bullshit singularity

#21

Post by horeszko »

opportune wrote: post #7

The internet is already mostly filled with low quality bullshit though, and GPT-4/Gemini are much better writers than whoever is churning out SEO as we know it now. It's a lazy argument to imply this 1. invalidates the technological achievement 2. prevents iterative improvement a la the singularity. For one, the Internet itself is not bullshit just because a lot of spammers/hustlers put bad content on it to try to ma…

Good content won't disappear necessarily, it could however be drowned out by BS as the author states.

Its the needle in a haystack problem, where AI is used to increase the size of the haystack, making the needle (i.e. quality content) harder to find.

Re: The AI bullshit singularity

#22

Post by dale_glass »

jsheard wrote: post #5

The same goes for image generation models, AI art already has a tendency to veer into the same clichés and those are only going to get reinforced if newer models are trained on newer scrapes which now include the million hyper-derivative AI images being uploaded to places like DeviantArt, Twitter and Pixiv every day. Those vendors who got in early have a moat in the form of untainted scrapes, but they'll eventually n…

Art is much less vulnerable.

Art is very heavily tagged, accurately described and filtered. Many galleries don't accept AI art at all, which means to sneak in the AI needs to be pretty much perfect.

Also, there's an enthusiastic scene of LoRAs where the makers work with small enough datasets to do manual curation.

Re: The AI bullshit singularity

#24

Post by jsheard »

rocqua wrote: post #13
jsheard wrote: post #5

The same goes for image generation models, AI art already has a tendency to veer into the same clichés and those are only going to get reinforced if newer models are trained on newer scrapes which now include the million hyper-derivative AI images being uploaded to places like DeviantArt, Twitter and Pixiv every day. Those vendors who got in early have a moat in the form of untainted scrapes, but they'll eventually n…

Are 'scrapes' really the only acceptable source?

Not always, a few companies like Getty Images are using their internal stock photo libraries to train models rather than slurping up the entire internet. They are uniquely positioned to take that path though, having already been in the stock photo business for decades before AI took off, and for everyone else it's a sisyphean task to manually curate a complete representative dataset that they can be sure isn't contaminated with AI images.

Re: The AI bullshit singularity

#25

Post by pixl97 »

rocqua wrote: post #9

Doesn't that just mean he who can get the best human feedback will get the best ai? If the quality of AI depends on not feeding on 'fake' input, the key becomes getting verified real input. My first thought was that this would create more money in genuine creativity, which would be great. But instead it feels more likely there will be much more telemetry and tracking to determine whether an input is made by you (and…

The Googles, Microsoft, Apples, and Amazon's of the world with access to 'real' humans are the ones with a leg up here. They already have the telemetry, now it's about effectively training models with it.

Re: The AI bullshit singularity

#26

Post by chmod775 »

smartmic wrote: post #2

I broadly agree with that. Repeated training with self-generated data is the technological equivalent of incest and can lead to nothing good.

Humans train on self-generated data all the time, but compared to today's LLMs, humans have superior reasoning ability, which enables them to separate good from bad data. Humans can confirm many things for themselves if need be and our social hierarchies crystallize authority figures which perform curation for others. On top of that humans have a lifetime of experience with the real world, while LLMs rely only on reports of what the real world is like.

All in all LLMs being more gullible than your average 6 year old is really not helping. Humans will have to (continue to) curate the data we train LLMs on.

Re: The AI bullshit singularity

#27

Post by frotaur »

ronald_raygun wrote: post #4

I always found the idea of infinitely self improving AI to be suspect. Let’s say we have a super smart AI with intelligence 1, and it uses all that to improve itself by 0.5. Then that new 1.5 uses itself to improve by 0.25. Then 0.125, etc etc. obviously it’s always increasing, but it’s not going to have the runaway effect people think.

Why should it be (1/2)*i improvement rate ? It could be 1/i ? I don't understand this argument.

Re: The AI bullshit singularity

#28

Post by rebolek »

opportune wrote: post #7

The internet is already mostly filled with low quality bullshit though, and GPT-4/Gemini are much better writers than whoever is churning out SEO as we know it now. It's a lazy argument to imply this 1. invalidates the technological achievement 2. prevents iterative improvement a la the singularity. For one, the Internet itself is not bullshit just because a lot of spammers/hustlers put bad content on it to try to ma…

If LLM needs so much babysitting then singularity is certainly not going to happen.

Re: The AI bullshit singularity

#29

Post by pixl97 »

rocqua wrote: post #23
gfodor wrote: post #6

Good cryptographic solutions exist for these problems

Cryptography can make content you create marked. But how would you use it to mark content other (non cooperative) people make?

Ya anyone that tells you crypto solves the problems of bad actors is delusional or pawning a bill of goods.

Re: The AI bullshit singularity

#30

Post by horeszko »

Needle in the haystack. Quality content is the needle, SEO spam is the hay. AI is the hay making machine.

Besides curation from a trusted source, what other solutions are there to prevent drowning in hay?

Post reply on HN