jj4211

joined 3 years ago
[–] jj4211@lemmy.world 2 points 14 hours ago

Note that it is mostly 'answered', but only really for the services and models themselves. Part of the defense in the case where Getty showed that it produced knockoffs of the Getty images was that "well, as the operator you are responsible for the output, not us".

So the operator is still on the hook if someone comes along and claims, even unwitting, infringement. In practice for code that's likely a tall order, as the sloperator is likely to keep their so close as to be a knock off pretty private, or the open source developer lacks the resources to realisticly track down offenders.

Closed source companies are more likely to come after folks, but that's a lower risk because they are so maniacly defensive about their code it probably never trained a model.

[–] jj4211@lemmy.world 1 points 14 hours ago (1 children)

Note that the remix/sample example hasn't always worked out as you state: https://en.wikipedia.org/wiki/Bitter_Sweet_Symphony#Credits_dispute https://en.wikipedia.org/wiki/My_Sweet_Lord#Copyright_infringement_suit

In many cases, "AFAIK" in your case you may have no idea that in fact, the copyright holder is being paid. Or the copyright holder is one and the same, with rights sometimes assigned to someone other than the musicians involved.

[–] jj4211@lemmy.world 2 points 14 hours ago (1 children)

So I had a huge refactor and thought "Ok, this should be right up GenAI alley". And in fact took your very approach of giving an example for a few and said "go at it".

To my surprise, it actually did it pretty poorly, would not work, when it would have worked, dire performance implications. Failing to address things that technically would survive the rework functionally intact, but now a very bad way of doing things in new context. Problems exacerbated is that when I'm reviewing code, I tend to have a more optimistic assumption of the code than when I'm writing and second guessing myself. So it's all the more annoying to read code I didn't write screw up so much.

I will say it does a pretty good job of boilerplate heavy crap. If I'm going to want to make Go structs from JSON, I can just feed a sample and the very tedious work of making the sometimes maddening tedius Go structs gets chewed through pretty nicely.

[–] jj4211@lemmy.world 1 points 14 hours ago

Oh man, the insane defensive code that it wants to generate. Yes, a decent principle, but when the stack has checked the same thing like 4 times, it's a bit much.

[–] jj4211@lemmy.world 2 points 14 hours ago

I've found 'senior' staff that became talking heads and stopped doing real work the bane of my existence in GenAI world. "I haven't coded in 20 years, but thanks to GenAI, I'm doing it again!" The reasons you stopped coding 20 years ago are plainly valid. Good for you, you got to transition to a grift based career where you say nothing but sound smart to the right people and get money, please don't return to coding because CodeGen is now 'cool'.

[–] jj4211@lemmy.world 2 points 14 hours ago

Have to think about this in an open source context. You have an open ended population ready to submit to your project if they think they can be useful

Now, thanks to GenAI, a bunch of people who aren't good at various things now thinks they are good at various things. Whether it's coding, making comics, making videos, what have you. Try to do nuance even if it strictly makes sense, and you are stuck with the slop problem if your project is sufficiently popular.

In terms of more directly on your points, if someone does GenAI to do a security analysis, ok, but I'd want them to do the tedium of trying to identify the false positives and then re-report it in their words of actual understanding. It can catch things, but along the way makes a haystack of sillyness to go through. Same for code review, legitimate issues, but lots of missing (I spent a non trivial amount of time yesterday because a GenAI code review insisted a variable would be unitialized when referenced, when I see an uncoditional assignment just a few lines above, trying to think if there was some catch I wasn't seeing). So indirectly using it and only subjecting the developer/maintainer to that which you know makes sense.

Problem being is that people have used Claude and have forwarded to me saying "I don't understand this well enough to judge, but forwarding to you just in case". I have the same tools doing the same things as you, I don't need meat proxies that just pass through the stuff without understanding.

[–] jj4211@lemmy.world 1 points 14 hours ago (1 children)

For school writing, it's more difficult to know, because the subject matter is so well trodden and students writing stuff they didn't really want to write trying to "impress" a teacher with wordiness and crap has a lot of the same vibe of GenAI writing. It's mostly obvious when there is a stark style/knowledge inconsistency with your personal knowledge of the student. Personal knowledge of a student is non-existent in lecture hall sized freshman level courses.

For code, well, at least the most problematic submissions are pretty blatanty obvious. They are obviously the result of a person asking for something nonsensical and the GenAI outputs content as consistent as possible with the stupid request, and a stupid request manifests in a very glaring way.

Sometimes it isn't the initial submission, but the resulting dialog that betrays it. Someone submits a small patch that is... well... short and to the point but the change doesn't seem to match the reported scenario it tries to address. In pursuit of clarification it becomes pretty obvious that the problem lacked sufficient actionable info, but the GenAI operator pushed it to produce something and out came something with a rationalization that sounds plausible but isn't anything.

Also the write up, of issues and code contributions. Humans are inclined to just make things to the point. GenAI sloperators make dissertations out of stupid simple things, with all sorts of tedious styling and crap. Frustrating because somewhere in the mess is their point, but it's buried beyond recognition.

[–] jj4211@lemmy.world 2 points 1 day ago* (last edited 1 day ago)

Have to dig into the nuance of this specific scenario.

A new memory vendor would be a huge capital expense, and investors are generally a bit apprehensive about that.

Further, it would be years before they could theoretically roll out product, a delay that investors would need to be awfully patient for under the best of circumstances. Further, we went through this dance in recent history, people thinking that the chip industry needed huge advancement and expansion of supply, only for demand to subside to normal before any of that expansion could even start.

And the stated payoff? Lower margin product than competition. Not exactly exciting to tell your investors your whole game plan is to make less money than your competition.

Then there's the reality that this is not an innate direct demand of memory for the sake of memory, it is intrinsically linked to these big AI companies, leading to the big question: Is this a bubble that has a risk of popping? If so, then the market will go poof before you have a single item shipped.

Even if broadly, you think the AI is viable, if any one company, especially OpenAI, gets left behind, the memory market could collapse. If not for Sam Altman's very specific purchasing commitments, the memory pressure would probably be much more modest.

Ok, fine, you are a ride or die believer in the durability of the AI boom and that every company is going to win. However, even if the AI companies do very well, what's to say they will still have the same appetite for hardware by the time this new enterprise gets going? A pivot from aggressive training to exploiting more what they have done, or some breakthrough that dramatically takes down their bloated memory requirements. If you believe in the AI boom, then just directly investing in the AI companies is the safer bet.

At the end of the day, an investor has a choice between being confident in the AI boom and investing directly in the AI companies, or being a bit less confident and investing in the memory vendors that are making bank now with a weaker, but still viable post-pop story. If you aren't comfortable directly investing in the AI companies now, then you almost certainly aren't comfortable with a long shot that only benefits if the AI boom keeps going exactly the way it has been going.

Yes, effort is underway to do this in China, but it's more about supply chain sovereignty than free market interests. It may have similar benefits, but here the free market is unlikely to be the impetus for increased supply in this scenario.

[–] jj4211@lemmy.world 2 points 1 day ago

SemVer isn't bad, but it's kind of pointless in the browsers. The value in SemVer is if you realistically promise to take bumping that first number seriously. Implying you take backwards compatibility seriously, and bugfixes seriously enough to keep patching an 'old' version. If you just maniacally bump the 'backwards incompatible change' number and never bother to revisit old releases, then I don't really care about the SemVer.

Of course, I also don't necessarily care about the CalVer either if there's update notification in play, the browsers will aggressively let you know you need an update. However if update notification isn't working or otherwise isn't in play, then CalVer can at least make you think "24.7... that seems like it might be old, maybe I should look for updates". In Windows world the CalVer has been informative as the system or corporate IT screw up has frozen a device at 22H2 and trigger some manual effort to figure out/fix whyever the hell the system won't go to new functional levels.

[–] jj4211@lemmy.world 13 points 1 day ago (4 children)

By definition, SemVer is supposed to have meaning to end users. If you see the third number increase, no worries, it's just bugfixes.

If you see the second number increase, well, in theory no worries, it's just cool new features but doesn't break anything you were doing, whatever you were doing should keep on working as always, and you can explore the new features at your leisure.

The first number changes: beware, something you may be used to can change/go away so it's not necessarily a slam dunk to update that.

The problem is that many projects just call it SemVer when they just play with arbitrary numbers. I'll call it "Marketing Versioning".

[–] jj4211@lemmy.world 22 points 1 day ago (1 children)

That's information superhighway thank you very much.

[–] jj4211@lemmy.world 24 points 1 day ago

Like the comment said, half.

view more: next ›