Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
-
@cwebber i can only read that statement as an outrageous troll. or at least i hope it's an outrageous troll, because otherwise it's the most misguided thing i've ever read about AI and programming. "we also dislike AI slop. which is why we use AI to generate choreography, not dancing" "which is why we use AI to generate sheet music, not musical performance" "which is why we use AI to generate blueprints, not buildings" "which is why we use AI to generate plots, not novels" etc
@aparrish @cwebber an emerging norm among people who believe they can save money by using LLMs is "AI use is okay as long as it's not noticeable in the final product". Have a human go back and fix the glitches that make your "placeholder" stable diffusion image, now it's "no longer slop". They notice that they only get yelled at when people can specifically detect the problem and conclude the issue is therefore whether the problem is *visible*.
-
@aparrish @cwebber an emerging norm among people who believe they can save money by using LLMs is "AI use is okay as long as it's not noticeable in the final product". Have a human go back and fix the glitches that make your "placeholder" stable diffusion image, now it's "no longer slop". They notice that they only get yelled at when people can specifically detect the problem and conclude the issue is therefore whether the problem is *visible*.
@aparrish @cwebber Recently saw an interview with a coder who got yelled at for using LLMs. The interviewer asked him what happened and he said "haha, it's very silly, at some point claude added a feature that has it co-sign your commits, that upset people so now I've turned it off". Then he goes on to talk about all the ways he uses claude. As if the problem were the commit header, or that he didn't hide it well enough, and not that *people are opposed to the use of the technology itself*
-
I think game developers especially (and I do game dev sometimes, sometimes even for work) tend to perceive code and art as interrelated and intertwined things. I find it unlikely that they can be easily separated.
I suppose some may see form vs function, but I personally see form *as* function.
@cwebber Slop is slop, at least for now.
This to me is like saying "We hate human trafficking, which is why we only traffic adults and not children!"
I genuinely don't think we can have an honest, critical conversation with most people about valid uses of machine learning and its offshoots until well after the bubble pops and LLMs become much harder to access and tech becomes much less interested in them and moves onto the next scammy tech.
Because despite how smart or aware people think they are or can be while using it, it isn't enough.
-
@aparrish @cwebber Recently saw an interview with a coder who got yelled at for using LLMs. The interviewer asked him what happened and he said "haha, it's very silly, at some point claude added a feature that has it co-sign your commits, that upset people so now I've turned it off". Then he goes on to talk about all the ways he uses claude. As if the problem were the commit header, or that he didn't hide it well enough, and not that *people are opposed to the use of the technology itself*
-
Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
> we also dislike AI slop. this is why we’re using AI to generate code, not content.
It's a philosophical distinction but one I feel like I don't get. Maybe it's because I like livecoding, etc, and see code itself as a form of art. Is AI code *not* slop in a way that feed content is?
And will vibecoded apps with Attie be likely to insert AIgen content?
To me that just feels even worse.
-
Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
> we also dislike AI slop. this is why we’re using AI to generate code, not content.
It's a philosophical distinction but one I feel like I don't get. Maybe it's because I like livecoding, etc, and see code itself as a form of art. Is AI code *not* slop in a way that feed content is?
And will vibecoded apps with Attie be likely to insert AIgen content?
@cwebber yeah I call bullshit. the artifact produced by slopcoding is absolutely slop.
-
@cwebber Paul's reply here has been living in my head.
Paul Frazee (@pfrazee.com)
As I said elsewhere I am basically radicalized about this. We fight now for personal computing and personal agency or we lose another decade to closed clouds We push now for an open internet and open models. Nobody is going to hand it to us because they’re nice
Bluesky Social (bsky.app)
I fundamentally can't understand this position. Pinning all your hopes for free and open computing on "open models," a thing that doesn't meaningfully exist, is so confusing to me.
But this does appear to be dogma for them.
@mttaggart paul is not someone i'd expect to make sense or be clever
@cwebber -
Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
> we also dislike AI slop. this is why we’re using AI to generate code, not content.
It's a philosophical distinction but one I feel like I don't get. Maybe it's because I like livecoding, etc, and see code itself as a form of art. Is AI code *not* slop in a way that feed content is?
And will vibecoded apps with Attie be likely to insert AIgen content?
It’s okay to view code as art, but it’s also okay to view code as a means to an end.
One can reasonably determine whether AI-generated code is sloppy or not through testing and review.
-
@wordshaper we have open models trained on generally-sorta-ethical data (BLOOM was one iirc), they're basically useless for the "open ended generation is all you need" applications
@mttaggart @cwebber@wordshaper they are very good as basis for focused nlp systems tho, but you can't get that funded anymore lol
@mttaggart @cwebber -
@mttaggart @cwebber We don't even *have* any open models as far as I'm aware, and short of someone sitting down with Project Gutenberg and maybe a copy of Wikipedia I can't see any way we'll get one for english text, and I'm pretty sure there's *no* properly licensed corpus of code for any programming language to do even minimal training there.
Every model I'm aware of is based on theft. (I'd love to be wrong, but that doesn't seem likely alas)
@wordshaper we have open models trained on generally-sorta-ethical data (BLOOM was one iirc), they're basically useless for the "open ended generation is all you need" applications
@mttaggart @cwebber -
-
@cwebber > … see code itself as a form of art.
When Knuth started his magnum opus about code, he very deliberately chose the title to be “The Art of Computer Programming”.
The Art of Computer Programming is named after Art Evans:
“I recall having lunch with a friend at the convention hotel. He knew how conceited I was, already at that time, so he asked if I was going to call my books "An Introduction to Don Knuth." I replied that, on the contrary, I was naming the books after him. His name: Art Evans. (The Art of Computer Programming, in person.)”
-
Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
> we also dislike AI slop. this is why we’re using AI to generate code, not content.
It's a philosophical distinction but one I feel like I don't get. Maybe it's because I like livecoding, etc, and see code itself as a form of art. Is AI code *not* slop in a way that feed content is?
And will vibecoded apps with Attie be likely to insert AIgen content?
@cwebber She is wrong, imho. All AI-generated code is slop, even when it "works".
-
@cwebber it's definitely art, but also, I think anybody that respects their craft, whether that be coding, music, writing, or anything else that takes human skill, would be ok handing some large chunk of it to generative AI
meanwhile, people that just want the result (including money)? usually the ones in favour of genAI
@OctaviaConAmore @cwebber It's also one where, like, I'm a writer, and I just absolutely hate the process of making all other forms of art, I can appreciate art, I just really don't want to do it. And yet the idea of using a diffusion model to add illustrations to my writing just has absolutely no appeal to me. It'd just make the story feel "cheap" and the models are far from able to be consistent. I feel like you just kinda have to not care about quality to be okay with anything GenAI
-
-
Been thinking about this: https://bsky.app/profile/jay.bsky.team/post/3micpg7z2h22g
> we also dislike AI slop. this is why we’re using AI to generate code, not content.
It's a philosophical distinction but one I feel like I don't get. Maybe it's because I like livecoding, etc, and see code itself as a form of art. Is AI code *not* slop in a way that feed content is?
And will vibecoded apps with Attie be likely to insert AIgen content?
regardless of what you're trying to make the quality of raw ingredients is paramount. substandard ingredients makes sub standard product. using a slop machine that is correct less than 50% of the time will yield slop product.
-
@cwebber certainly though I think the more you abstract yourself from the code the more potential for problems there is. If the tool is writing code in your editor and it produces something that's 95% of the way there, you can just go and tweak the last 5%. If it's sending PRs straight to GitHub or whatever, you're much more likely to let that little bit slide because it's so much more effort to fix. And if you're not looking at all, well, you're running blind
"toil" is a great description!
-
@cwebber Paul's reply here has been living in my head.
Paul Frazee (@pfrazee.com)
As I said elsewhere I am basically radicalized about this. We fight now for personal computing and personal agency or we lose another decade to closed clouds We push now for an open internet and open models. Nobody is going to hand it to us because they’re nice
Bluesky Social (bsky.app)
I fundamentally can't understand this position. Pinning all your hopes for free and open computing on "open models," a thing that doesn't meaningfully exist, is so confusing to me.
But this does appear to be dogma for them.
@mttaggart @cwebber open models do exist and I've been kicking the tires with a few but have not attempted to use any of the ones for code generation. They are definitely less sophisticated than the top of the line LLMs from what I have seen but that is unsurprising given the usual comparison between scrappy FOSS and highly resourced proprietary software https://osai-index.eu/database/
-
@mttaggart @cwebber open models do exist and I've been kicking the tires with a few but have not attempted to use any of the ones for code generation. They are definitely less sophisticated than the top of the line LLMs from what I have seen but that is unsurprising given the usual comparison between scrappy FOSS and highly resourced proprietary software https://osai-index.eu/database/
@ehashman @mttaggart @cwebber there are some capable open models. My poorly documented journey is here
I’ve been too busy lately to do anything interesting, but it’s pretty fun learning
-
@cwebber Paul's reply here has been living in my head.
Paul Frazee (@pfrazee.com)
As I said elsewhere I am basically radicalized about this. We fight now for personal computing and personal agency or we lose another decade to closed clouds We push now for an open internet and open models. Nobody is going to hand it to us because they’re nice
Bluesky Social (bsky.app)
I fundamentally can't understand this position. Pinning all your hopes for free and open computing on "open models," a thing that doesn't meaningfully exist, is so confusing to me.
But this does appear to be dogma for them.
@mttaggart @cwebber Having absolutely disagreed with Paul over AT early on in BlueSky, I am not surprised to see that he appears to be saying all the right things whilst, again, talking about a closed system (this time AI).
Sometimes I wonder if people's experiences are just so skewed that words are being redefined.
