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

Things are about to get worse for generative AI

garymarcus.substack.com

Post reply on HN
Print view View original

251–260 of 769 posts

Re: Things are about to get worse for generative AI

#251

Post by ben_w »

sensanaty wrote: post #228
ben_w wrote: post #207

Earlier quoted context omitted.

> they're not even close to being AGI Only because all three letters of the initialism mean different things to different people. Existing LLMs won't do everything, but bluntly: good, we're not ready for a world where there is an AI that can do everything for $1-60/million words[0], and we need to get ready for that world before we find ourselves living in it. ChatGPT-3.5 has a lot of weaknesses, but it can still do…

> ChatGPT-3.5 has a lot of weaknesses, but it can still do a better job of coding than a few of my coworkers demonstrated over the last 20 years. Whenever people say stuff like this I can't help but wonder what on earth kind of projects they work on. Even GPT4, while useful for things like reformatting or generating boilerplate code and stuff like that, it's still a far cry from any decent dev I've ever worked with,…

> Whenever people say stuff like this I can't help but wonder what on earth kind of projects they work on.

Terrifyingly, one of the bad human examples was doing C++. That person didn't know, or care to learn about, the standard template library; and they also duplicated entire files rather than changing access specifiers from private to public so they could subclass; and one feature they worked on was to support a change from storing data as a custom file format to a database, and the transition could take 20 minutes on some inputs even though neither loading before nor after this transition took more than milliseconds, and they insisted during one of the standups the code couldn't possibly be improved… the next day I looked at it for a bit, removed an unnecessary O(n^2) operation, and the transition code went back down to milliseconds. Oh, and a thousand(!) line long block for an if statement that always evaluated true.

The whole codebase was several times too big to fit into the context window for any version of any GPT model thanks to both this duplication and to keeping old versions of functions around "for reference" (their words), but if it had been rewritten to be more sensible it might just about fit into the biggest.

(My other examples were either still at, or fresh out of, university; but this person should have known better).

> Don't even get me started on the actual difficult part which is the whole preamble to creating the ticket in JIRA or whatever task management software you use where you're talking with stakeholders and planning out the work ahead, you're telling me you're paying 'Open'AI to do that whole rigamarole for you, and you're doing it successfully?

If it was all-round good, none of us would have jobs any more.

Re: Things are about to get worse for generative AI

#252

Post by pier25 »

Baldbvrhunter wrote: post #2

I imagine the argument might be like this: I hire a session musician to play on my new single, paying him $100. I record the whole session. I ask him to play the opening to "Stairway to Heaven" and he does so. "Well, I can't use that as a sample without paying" "Ok play something like Jimmy Page" "Hmm, still sounds like Stairway to Heaven" "Ok, try and sound less like Stairway to Heaven but in that style" "Great, I'l…

The guitarist is not publishing the content, you are.

It could be argued ChatGPT is a publisher too.

Re: Things are about to get worse for generative AI

#253

Post by atq2119 »

keiferski wrote: post #127
bnralt wrote: post #124

Earlier quoted context omitted.

It seems like a somewhat dystopian thing to fix. Imagine a scenario where Photoshop would scan images you uploaded for copyright material and then refuse to work if it determined image contained any copyrighted material or characters (even if it was just a fan drawing you did). This reminds me of the early days of the internet where people wanted to remove free fanfiction for violations of copyright laws. Trying to a…

I doubt that the media companies would be happy about this; but maybe a compromise is a “copyright infringement filter” that can be enabled or disabled, with a flashing notification that you’re responsible legally if you turn off the filter and have issues.

Sure, there are legitimate but opposing interests here. The solution doesn't have to be technical though. The key part is making sure that the copyright owners still have some recourse, but one that isn't punitive for unknowing infringement. For example, make it legally impossible to impose punitive penalties for unknowing infringement without commercial interest, but make it possible for the rights owner to demand the relevant material be removed etc.

Also keep in mind that the Mario examples from the article are perhaps not the best guide here. Mario is sufficiently pervasive in our culture that you can't reasonably claim unknowing infringement. It's the somewhat more obscure cases that I'd be worried about.

Re: Things are about to get worse for generative AI

#254

Post by leereeves »

pama wrote: post #241
dkjaudyeqooe wrote: post #111

Earlier quoted context omitted.

> a few copyright holders By which you mean every copyright holder. > AGI in the near future Something that is purely speculative, undefined, and has been promised in the near future for 50+ years. I don't see copyright holders lying down for someone else's benefit and I don't see governments gutting copyright, contract law, and several other avenues of protection that copyright holders can deploy in the name of some…

Japan is already taking action regarding copyright law and the AI world is noticing: https://www.natlawreview.com/article/japanese-government-ide... https://www.cliffordchance.com/insights/resources/blogs/talk...

Neither of those sources describe any actions "gutting copyright law". They just say that officials met to discuss the issues.

Re: Things are about to get worse for generative AI

#255

Post by throwoutway »

redcobra762 wrote: post #248

This operates similarly to importing an image into Photoshop. You can do whatever you like with images privately, or with gen AI, but the game ends when you try to use those images commercially. Not sure how this “gets worse” or better for anyone. The current state of things seems generally fine, and there’s a real possibility the courts see it that way too.

> but the game ends when you try to use those images commercially.

Right now, it feels more like it's called "innovation" and "entrepreneurship" than the end-game, as long as you have billions invested. Waiting on the courts to decide this issue

Re: Things are about to get worse for generative AI

#256

Post by _xnmw »

Havoc wrote: post #55

To me that’s the wrong question. Everyone knew it was trained on copyrighted material and capable of eerily similar outputs. But it’s already done. At scale. Large corps committing fully. There is no chance of that toothpaste going back in the tube. It’s a bit like when big tech built on aggressive user data harvesting. Whether it’s right, ethical or even legal is academic at this stage. They just did it - effectivel…

So you're saying this is a fait accompli. Like many great innovations in tech, break the law because the law is silly; remember when Uber and AirBnB were illegal in most major cities and achieved market dominance anyway?

I say, good riddance. I never believed in any such thing as "intellectual property" anyway, I say, get rid of it all, patents, copyright, and the whole pile of imaginary "rights". More than half the world (i.e. the Global South) don't recognize these rights anyway, and it is becoming increasingly difficult to enforce it without draconian legal overreach and monopolistic centralization.

Re: Things are about to get worse for generative AI

#257

Post by numpad0 »

gchamonlive wrote: post #62
keiferski wrote: post #39

These don't seem all that difficult to fix to me. Most of the examples are not really generic, but are shorthand descriptions of well-known entities. "Video game plumber" is practically synonymous with "Mario" and anyone that has the slightest familiarity with the character knows this. Likewise, how difficult is it to just use descriptive tools to describe Mario-like images [1] and then remove these results from anyo…

The thing is that those are really trivial or extreme examples. What we should take from this: 1. Generative AI systems are fully capable of producing materials that infringe on copyright. 2. They do not inform users when they do so. So potentially any output could be infringing copyright source material, even from some obscure but still protected corner of the web, and anyone using that output could be exposed to la…

AIUI/IIUC/IANAL, Generative AI systems stack up tuned butterfly effects to create meaningful outputs. Which means, data is not merely stored inside a butterfly but spread across the system, and potentially every output from the butterfly cage is infringing everything, just almost negligibly. But no one can prove that or otherwise to be the case with current technology.

I think it's just unfixable as far as fixing goes. The "best" way is to nuke all GPTs and DALL-Es and bury the technology, and the next best is to mark them all un-copyrightable and un-publishable as a compromise. This second option should be an all-around win that also encourages edge and on-prem deployments, IMO.

Re: Things are about to get worse for generative AI

#258

Post by Baldbvrhunter »

atq2119 wrote: post #171
kredd wrote: post #7

Earlier quoted context omitted.

You, because you released the song and took the royalties? I don’t think every type of art can be compared against each other though, as there have been numerous precedents specifically for music, some for paintings, and some for photography with their own nuances. I still think people who are concerned that art related copyright will stifle generative AI should fight copyright laws directly. But that’s a harder pill…

Part of what's interesting here is that generative AI makes it very easy to unknowingly and unintentionally get on the wrong side of copyright law, which is something that wasn't really possible before. That's something which, IMHO, should be acknowledged by the law.

Ask George Harrison about "My Sweet Lord" which cost him $587,000 for his unconcious infringement.

Re: Things are about to get worse for generative AI

#259

Post by andsoitis »

d4mi3n wrote: post #161
squidbeak wrote: post #142

Earlier quoted context omitted.

If a child is instructed to read a copyrighted work at school, which later becomes a factor in his own derivative works, he won't be in breach of copyright. Why should other intelligent entities be prevented from reading copyrighted works and gaining whatever there is to gain from those works the way any human might?

This argument might hold more water when generative models are more than fancy compression algorithms/text completion engines. A more practical way of looking at this is: who is making money off of these models? How did they get their training data? I’m not a fan of copyright in general, but we have serious outstanding issues with companies and organizations stealing or plastering work without compensating the origin…

> I’m not a fan of copyright in general, but we have serious outstanding issues with companies and organizations stealing or plastering work without compensating the original creators of said works

Copyright is meant to give the original creator a monopoly over their creation (so that others don't profit off of their work). Are you not a fan of copyright in its current scope / implementation? Because it sounds like you do agree with its goal.

Re: Things are about to get worse for generative AI

#260

Post by cogman10 »

josh-sematic wrote: post #223

Gary Marcus is growing his subscriber base using images of copyrighted IP (C3PO, Mario, etc.). Fair use? Then why is the tool he used to produce those materials not also fair use of the IP? My take is that either we say the models are like people (do we penalize people for learning from IP and letting that influence what they subsequently produce?) or we say they are like tools (do we penalize Adobe because Photoshop…

Because the fair use clause he's using is about giving commentary.

The reason the tool is problematic is because derivative works are also copyrighted. LLMs aren't adding value to their output or using creative functionality. That are smashing multiple works together to produce a response. And, many of them are selling the output which is doubly problematic.

Consider this, if I sell a book about gandolf and Dumbledore getting into a wizards duel, both jk and Tolkien have grounds sue me. Adding another copyrighted source does not protect me.

This is especially a big problem in the music industry.

Now should copyrights be like this? I don't know. It feels to me that copyrights have the wrong balance all over the place.

Post reply on HN