Christoph WieseFeb 16
James Thomson

Writers: Generative AI models were built on our stolen works, are deeply unethical, and risk devaluing our entire profession.

Artists: Generative AI models were built on our stolen works, are deeply unethical, and risk devaluing our entire profession.

Developers: Wheeeeeeeeee!

Feb 16 at 12:40amIvory for Mac
Show threadJonathan PolleyFeb 16
@jamesthomson I stopped supporting ATP because of their change from “AI is theft” to “you should pay $20 per month for ChatGPT.”
Show threadJared White (ResistanceNet ✊)Feb 16
@the_other_jon @jamesthomson John Siracusa seems to offer the most balanced perspective there, but yeah it's pretty grim. I mostly don't listen any more. And don't get me started on MacStories omfg 😵‍💫
Show threadAlexandraFeb 16
@the_other_jon @jamesthomson I was confused about this change as well. But they use Apple products, while being sceptical about the company. What’s wrong about using ai as a useful tool, while knowing its problems?
Show threadJonathan PolleyFeb 16

@owlex Are the training sets licensed or just strip mined from the web/redit/github/sourceforge? This was the cause for their “AI is theft” statement.

From a technical standpoint: Are these training sets free from bugs? If you use an ai tool to generate tests, are they useful tests? A useful test is one that tries to break the code instead of showing that the code “works. Tests that that exercise the interfaces or cover the code tend to not be “useful” tests.

Show threadAlexandraFeb 16

@the_other_jon I'm aware of these problems, and many more (energy waste, OpenAI's exploitation of workers in Africa for manual training, copyright theft, data mining).

My question stands: Why is it wrong to use something critically while being aware of its problems? Especially when we're in the middle of such a massive technological shift that we should understand it. And when capitalism is forcing it into everything anyway, isn't informed usage better than ignorance?

And it's not even just about American companies anymore. We're in a global race for AI dominance now. This whole topic is incredibly complex.

I respect you for having these principles, but I think taking it out on a podcast, which reports about technology is a little weird. Though it's your decision 😊

Show threadAiris DamonFeb 16
@owlex @the_other_jon If it is incredibly complex, then shouldn't the technology be democratically controlled? Shouldn't all tech that has such a massive impact on our lives be democratically controlled? I believe it should.
Show threadAlexandraFeb 16

@airisdamon @the_other_jon

I'm not sure we need to democratically control the technology itself, but we absolutely need to hold companies accountable for their methods. And since these models are built on OUR collective knowledge, we should demand open weight models and not be forbidden from using them.

The true impact of LLMs is still unfolding. If they turn out to be like the telephone or internet, then yes, strong regulatory control is needed. But if they're more like one compiler among many, maybe not.

What's clear: We need to close the legal loopholes that let companies profit parasitically from society without giving back. Democratic control means informed engagement, not avoidance.

Show threadillogicalFeb 16

@owlex @airisdamon @the_other_jon

One criticism on your choice of words: large language models are not built on knowledge, but on data.

I think that is a very crucial distinction to keep in perspective what the tools can and cannot do.

Show threadJohnFeb 16
@airisdamon @owlex @the_other_jon It's not gonna fly. Apple doesn't release their source code. People still pay them money for some reason. Knowing what the code does is an infinitely easier step (and a prerequisite to) controlling what code does via legislation. It doesn't matter what 'society should do'. Society will keep paying Apple. Apple will keep paying government to make sure it's never compelled to reveal what its code does to its users.
Show threadAiris DamonFeb 16
@mrkeen @owlex @the_other_jon There could be extralegal methods for democratization. I don't know. I'm not real enthused with the direction all this is heading toward.
Show threadStephen 🌈 (he/him)Feb 16

@owlex @the_other_jon An ethical position on something often requires sacrifice. We aren’t doing this to be mean to the podcast. We are doing it to attempt to influence the industry in another direction.

The complexity of the situation doesn’t really have anything directly to do with what is ethical. It only has to do with how hard it is to see it. Are you arguing that the complexity makes it ok or that it is hard for you to see? Some of us can see the harm and are trying our best to make it visible.

Those who provide the counterpoint don’t say anything about whether the harm will stop or somehow be mitigated really — they mostly just say, “Don’t be left behind.” Does that sound like a rational actor or an addict?

My belief: it is absolutely wrong to feed this technological vampire that threatens to erase humanity. Don’t become a thrall. It doesn’t end well for them. 😊

Show threadAlexandraFeb 16

@firepoet @the_other_jon

Thank you for this answer 😊

I respect that position, and you're right that ethical stances often require sacrifice. But I think we're drawing different lines here. I don't see AI as something we can just starve by not using it. It's already everywhere, being used by corporations, governments, everyone. So for me, the question is: do I abstain entirely while others use it uncritically, or do I use it thoughtfully and keep pushing for better regulation?

I am trying to navigate the reality I am living in.
As I said in other posts it helps me navigate with my ADHD and also I’m kind of forced to use it at work.
What I am doing is to advocate for using different models at work and talking with people about the problems of AI. All while being really excited how we can use it to make lives better, because it has cool use cases.

Also I am pragmatic singular boycott never helped much. We need regulations in place, that’s the most important thing

Show threadStephen 🌈 (he/him)Feb 16
@owlex @the_other_jon I wish you the best. While we are on opposite sides of this particular struggle I can respect your need to fix the broken system from the inside. Just be careful out there.. https://steve-yegge.medium.com/the-ai-vampire-eda6e4f07163
The AI Vampire

This was an unusually hard post to write, because it flies in the face of everything else going on.

Medium
Show threadAlexandraFeb 17
@firepoet Thank you, Stephen. While we still don’t have the same position. This post captures what I am looking out for in my employees and myself
Show threadMichelle AndersFeb 23
@owlex @the_other_jon great page you’ve got Alex ✌️
Show threadFelipe CeprianoFeb 17
@the_other_jon @owlex uh, I definitely disagree on your stance about tests: A test that checks if something is working is really useful when you need to refactor something and needs to be sure the changes haven't affected existing behaviour
Show threadJonathan PolleyFeb 21
@felipecn i have run into more than a few instances where the software didn’t really work. The testing passed because they tested what worked and not what it was supposed to do.
Show threadkaffeebär 🐻Feb 16

@owlex
What's wrong about driving (ICE) cars? What's wrong about flying? What's bad about cruise ships?
Such great tools to get from a to b.

@the_other_jon @jamesthomson

Show threadMatt MassicotteFeb 16
@jamesthomson how dare you
Show threadAmy ParentFeb 16
@jamesthomson the amount of podcasts I listen to is dwindling at this rate. The amount of “yeah but look Claude my best pal made an app for me so it’s all good” is making my blood boil
Show threadJames ThomsonFeb 16
@amyinorbit I mean, aside from the personal, ethical, societal, financial, and environmental issues, it's just great.
Show threadChris Mackay 🇨🇦Feb 16
@jamesthomson @amyinorbit I’ve dropped ATP, MPU, Connected, MacStories, Intelligent Machines, & Windows Weekly. The worst of them have actually scolded listeners who disagree with them, while others have made it clear that we’re old, clueless technophobes 🙄. I have my eye on others than tend in the direction of having drunk the AI koolaid. The only thing that’s inevitable about all this? I may never have to skip another fucking ad for SpareSquace.
Show threadCarolynFeb 16

@tantramar @jamesthomson @amyinorbit “actually scolded listeners who disagree with them, while others have made it clear that we’re old, clueless technophobes…”

Basically doing what those in the US government are doing: instead of addressing criticism, distract and avoid addressing critiques or questions by attacking or bringing irrelevant things into a conversation.

Show threadMathias PanzenböckFeb 16
@CStamp @tantramar @jamesthomson @amyinorbit The last Corecursive episode was strange. He basically described how he's overworking himself managing coding assistants, but he's excited on how much code he's generating. At least he's not talking down on people not using AI. I personally want to be a programmer, not a manager. It feels like all these people didn't really like programming?
Show threadFabienFeb 16

Related: I’m looking for suggestions to add more diversity to my tech podcasts rotation, something different than 3 middle age white US dudes with enough money to buy Porsches.
(still love you @atpfm)

/cc
@amyinorbit @jamesthomson

Show threadsoftmausFeb 16
@fabienmarry @amyinorbit @jamesthomson It has always been mind boggling to me how they mock Casey for being ‚frugal‘ and how it’s ‚funny‘ that Marco apparently has some form of a money spending disorder.
Show threadFabienFeb 16
@softmaus It's all relative isn't it… Frugal means thinking of replacing a top of the line laptop because it's 2 year old… while the other host buys a restaurant. I still find that interesting, but less and less relatable.
Show threadsoftmausFeb 16
@fabienmarry Same. In the meantime, I manage to cut them some slack by remembering that consumerism still is a commonly accepted virtue in the US.
Show threadxenonchromaticFeb 16
@jamesthomson "risk devaluing our entire profession..." To the people developing these models, this isn't a bug, its intended behavior.
Show threadJames ThomsonFeb 16
@xenonchromatic Indeed.
Show threadThomas BrandFeb 16
@jamesthomson DragThing now by ChatGPT.
Show threadmacfixerFeb 16
@Eggfreckles @jamesthomson
Show threadDavid H. McCoyFeb 16

@jamesthomson

I know a manager who suggested someone use AI to write a simple description of a piece of code.

Nuts.

Show threadColin CornabyFeb 16
@jamesthomson I think the problem is developers don't really consider any unattributed use of open source as stealing - just a mild grey area. (They should consider it stealing.)
Show threadJared White (ResistanceNet ✊)Feb 16

@colincornaby @jamesthomson Open source != public domain, and free software != free (it's free speech, not free beer), but apparently many developers are clueless re: all those nuances. ☹️

Perhaps if all LLM-generated code was legally automatically placed in public domain, we'd see a bit of a light bulb moment. 😂

Show threadColin CornabyFeb 16
@jaredwhite @jamesthomson All LLM generated code is in the public domain. The commercial companies just protect it all behind private repos. If you could force them to release it that would be what you’d need.
Show threadUr Ya'arFeb 17

@colincornaby
@jaredwhite @jamesthomson

A post suggesting precisely this:
https://zomglol.wtf/@jamie/116059523957674208

Jamie Gaskins (@[email protected])

Attached: 2 images If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*. This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain. Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

zomglol
Show threadDavid W. JonesFeb 16
@jamesthomson Also developers: Vibe-coding generative AI models were built on our stolen source code, we're all being laid off, please hire us.
Show threadMark FischerFeb 16
@jamesthomson
I would be interested in training a model based on my own code. I spend a decent amount of time looking through my own code to find something I know I’ve done before.
Show threadJames ThomsonFeb 16
@estranged My understanding is that there isn't enough data in a small set like that to actually train usefully, as a standalone thing. So it's always going to be based on other models, with your training on top.
Show threadJack BrewsterFeb 16

@jamesthomson @estranged This is something I've wondered about. I hadn't considered small data sets would be an issue.

Seems like a solvable problem though. There's probably not a business (i.e. profitable) for it though. Yet.

Show threadMathias PanzenböckFeb 16
@estranged @jamesthomson So you want better semantic search. That I also want.
Show threadMichael B. JohnsonFeb 16
@jamesthomson I’ve been struggling with this cognitive dissonance for years.
Show threadJames ThomsonFeb 16
@Drwave Let me know if you come to any conclusions…
Show threadGreg PierceFeb 16

@jamesthomson @Drwave If you put an appetizing plate of food in front of them, most people won’t ask to look in the kitchen of the restaurant, and they will actively avoid the subject if you try to tell them how the meat was raised.

I'm not defending these tendencies, but they are not different.

Show threadJames ThomsonFeb 16
@agiletortoise @Drwave This meat tastes suspiciously familiar.
Show threadGreg PierceFeb 16
@jamesthomson @Drwave
Show threadMichael B. JohnsonFeb 16
@agiletortoise @jamesthomson Ah, Greg - found the GIF before me…
Show threadMichael B. JohnsonFeb 16
@jamesthomson @agiletortoise I don't want to know what Soylent Green is, James…
Show threadDaniel GibsonFeb 16
@jamesthomson @agiletortoise @Drwave
They say music is the food of love
Let's see if you're hungry enough
Take a bite, take another
Just like a good boy would
Get a sweet thing on the side
Home cooking, homicide
Side order, could be your daughter
Finger licking good
Show threadwFeb 16
I'm torn whether this is cognitive dissonance or a result of an industry containing a significant fraction of people who apparently hate the field they work in

CC: @[email protected]
Show threadLona TheartlavFeb 16

@jamesthomson I feel like most devs don't see code as art, but as something ephemeral and disposable, a mean to an end.

Like i quit IT years ago and mostly code for my own use, and i'm closer to writers and artists on this one.

Show threadParsingphaseFeb 16
@jamesthomson I touched on this earlier… https://m.phase.org/@parsingphase/116076869609984951 & linked
Show threadJames ThomsonFeb 16
@parsingphase Very true!
Show threadLee from ColoradoFeb 16
@jamesthomson honestly it seems best for coding...
Show thread[email protected]Feb 16
@colo_lee @jamesthomson people that know jackshit about coding would say that...
Show threadCarlos SolísFeb 16
Developers assume that all that code on the Internet must probably be open-source anyways. Cluelessly ignoring issues of attribution requirements, license compatibility, and the sheer lack of tracking of either by most scrapers. So far I've found one single model that even attempts to keep tabs on that, so you can imagine the ticking legal bomb that is the rest of them
Show threadTravis MarbleFeb 16
@jamesthomson It is depressing, I think some of it is management push, I think most of it is FOMO. Developers have always had this unique personality, they have to be seen as early adopters, innovators, and I think pontificating about how good they are at using AI, is just that. Just look at LinkedIn.
Show threadverseFeb 16
@jamesthomson man this made me laugh so hard. Mastodon is cool