Worth repeating: Claims of "AI" democratizing anything (coding, creative endeavors, etc) are always BS. The thing locking out people from doing that is not having the time/resources.

You want to democratize coding/art/creativity? Give people paid time off to do it and access to infrastructure. Easy.

I do want to see the art people come up with who can't do it today. Give everyone a paid month off every year to do whatever they want. Learn a thing, make music. Paint. Anything.
That gives people access. Not a slop machine.

@tante
The slop / plagiarising machine gives giant corporations control,

Also you can't copyright / own any AI generated content. You can be sure if it could be copyrighted, the Corp selling the service would own the copyright.

@tante I thing they mean "we can pay *anyone* to attempt to fix the crap the LLM excretes, therefore the salary is whatever is the least any person makes right now"

@tante

Surely there's no way that hitching my wagon to one of the three giant world-eating dragons would in any way limit my freedom of movement and choice of direction?

@tante Furthermore: People are not creating, they are commissioning.

"I used Da Vinci to create the Mona Lisa." - some Medici, probably.

(I'm not entirely sure about the Da Vinci- Medici - Mona Lisa connection. Feel free to give me a more historically accurate example. And feel free to use it in discussion about AI "art". :) )

@DerGiga The Mona Lisa pictures the noblewoman Lisa del Giocondo. The portrait was commissioned by her husband, Francesco di Bartolomeo di Zanobi del Giocondo.

https://en.wikipedia.org/wiki/Lisa_del_Giocondo

Lisa del Giocondo - Wikipedia

@DerGiga Pope Julius II used Michelangelo to decorate the ceiling of the Sistine Chapel
@Anke Thanks, that would be understood as well and it's historically accurate. <3
I should use that in the future.

@DerGiga @tante and every artist who had to wrestle their way through a client’s near-inscrutable ‘prompt’ to arrive at a competent picture knows very very intimately that natural language is a really bad interface for such a purpose, too.

this weird dream of ‘i tell the machine what i want and then i get the picture i had in my head’ is making some very bold assumptions about how well they can wield language.

@tante It ain't democracy, it's dictatorship. A sloptator (wait that sounds delicious...)

Democracy is a visibly collective effort. Everyone has to participate for it to really work. That's the opposite of what AI does, AI absorbs that collective effort and spits it out as its own individual achievement. Like a dictator or a CEO claiming credit for the entire country or company.

@tante what always gets me: it's not "democratizing", it's "commoditizing"
@tante If anybody trying to sell me tech says anything about "democratizing [thing]", I assume they have bad intentions. You're not democratising shit. You're trying to build vendor lock-in and pretending you're motivated by altruism
@tante I agree. And yet there is this: I have been using small LLMs to help me kick-off coding projects that I conceptually understand but lack the mental ability to learn the programming on my own. I let the model sketch the structure, suggest a stack, and then let it build while I watch and ask questions. Then I tell it to document all steps and explain what does what on a granular level. Last week I was able to generate a simple SSG. (1/5)
For years I have been structuring it in my mind. It turned out pretty functional. Believe me, no time and dedication would have had the same empowering effect in the same time frame. I am a really slow learner when it comes to the concrete programming languages and their topography. This has been holding me back the last 15 years. Now I can watch functional code being created and see what it does or does not. And what do you think I do? Next time I go and copy the project by hand. (2/5)
Learning how to do it in the process. It's maybe a niche approach, but it works. I only rely on code generation as long as absolutely necessary and not out of comfort. (No "center the div pls"). In a way, the model slowly makes itself obsolete. Another principle I have is to only let code be generated that I understand. Once I don't know what it does, I stop. And the third principle is to not do any potentially harmful coding projects this way. (3/5)

I code small websites and automations for silly podcasting and conversion of files. That's it.

I don't see this niche talked about much. I have an attention disability. LLMs help me. Believe it or not. (4/5)

You referred to home assistant on another post doing some of the same stuff OpenClaw does. In this context I agree. The equivalent here would be an interactive programming tutorial. They have existed before wide adoption of LLMs, yes. But they were never able to adopt to real context. Instead they are usually prepared and edited by humans as a constrained lesson. Usually there is no room for "what if we did this?". (5/5)

@eurodivergent @tante this is exactly why I stand by #llms democratizing access to software engineering. I know so many people know, from artists to blue collar to family business hustlers to genomics researcher who are now building full fledged software solving problems for them. I assist minimally. They are truly empowered and so much of the alienating side of tech is gone, from replacing photoshop to using local only html gizmos to just knowing that you can get a relay board connected to your music software without spending the whole weekend on arduino frustrations. It cuts off ties to big tech and gets us back to whimsical personal software.

I can’t speak to other domains, but this one is truly real.

@mnl @eurodivergent @tante I think people tend to conflate output with process and semantics confuse us.

If you "commission" an AI as a layperson to write code, you did not code an app or know how the thing works and how to fix issues. Nor did you "create" art by telling a stochastic black box to approximate an output commensurate with what you would have created, if only you had the wherewithal to do so.

Commodification of the output is true, as long as the gatekeeper allows you to afford it.

@jakob @eurodivergent @tante but if I commission a computer to printf(“hello world”), didn’t I just commission an entire tower of abstraction and code and silicon and engineering to print out hello world? Where do you draw the boundary between “print hello world” and “printf(hello world)”
@jakob @mnl @tante Nothing done with computers is solely a human thought process. Or to invert the same argument: Every human-computer interaction consists of dialogue. Machine output is the only possible way of feedback. The very reason computers exist is to fold complex actions and free us from doing them manually. I'd draw the line not between process vs. output but rather between understanding and not understanding what the computer does.

@eurodivergent @jakob @tante while i don't fully subscribe to computers being purely utilitarian, I agree that how much understanding you want to put into what areas is something each person gets to determine for themselves, and doing things with a computer / understanding what a computer does is not just about coding (and even in coding, there are so many layers. I thought that in order to write printf("hello world") you had to learn how to build an OS and device drivers and how to build a CPU and at some realized that it's ok to rely on preexisting software and not understand everything).

It still gets me because I can't for the life of me not take things like "to be a real programmer you have to understand your code does" literally, because as hard as I try it is just absolutely impossible. At this point I don't even have a reasonable mental model of how the smallest microcontroller work anymore, with them coming with 1200 page datasheets. How am I supposed to understand what my javascript actually does?

But I don't think anybody writing javascript actually does so thinking about even something as high level as assembly instructions (except the odd JIT/interpreter implementer, and even then...).

So I do think it's about the output primarily, because a good output (i.e. something of value to humans) usually requires a lot of care, iteration and going beyond code.

@eurodivergent @tante

You need to do a real course. You are simply being exploited.

@raymaccarthy @tante How am I being exploited? I learn how to build a template system while taking a shit. Also I do courses and will continue to. What I described is just a parallel approach that happens to work for my disability.
@raymaccarthy @eurodivergent @tante You need to stop telling people who didn't ask your opinion what to do. You're being annoying.
@tante we’re democratizing coding! That will be 20$ per month, please

@tante I spy with my little eye a connection between this sentiment of "AI democratizing things" and UBI not being implemented as policy 👁

Spot-on 👏

@tante Also: modern art has thoroughly democratized art already. A fake name signed to an urinal is considered art. What's left to democratize?

@tante this is an intellectually dishonest argument.

Everything is solvable by "more time and more access to infrastructure" when your limits approach infinity, similar to how, if you gave an infinite amount of monkey an infinite amount of typewriters, one of them will eventually type up with the works Shakespeare.

Take creating an alternative to Android for example, it's a Herculean task not because of the OS itself, but because of a

> wide range of nonfree software blobs that commonly occur in even the most progressive "free software" operating systems.

https://librephone.fsf.org

LLMs are great at pattern recognition by design, including identifying function signatures, recognising common driver patterns, reading Ghidra output, etc.

This takes a problem that was so far considered in the realm of needing state backing to a problem that can be tackled by a relatively small team of dedicated researchers.

So yeah, my bad if I don't give a damn about moralising "AI use" when the alternative is rewriting and drying up the moat of American tech companies that enable surveillance and brutalising of the Global South

LibrePhone

LibrePhone is a project to research freely licensed firmware for mobile phones.

LibrePhone
@budududuroiu @tante LLMs start from the intellectually dishonest position of reselling stolen goods.
@DataGhost @tante love to see it, Mastodon arrived at the "you wouldn't download a car" argument. Slow clap lmao

@budududuroiu @tante Whatever gets you through the night.

The LLMs are abusing the generosity of the Free and Open Software movements to destroy them by commoditizing the work of the Community.

No matter how noble the goal, these beginnings are unclean.

@DataGhost @tante where is the abuse of generosity if you train and distribute a GPL-licensed LLM? Chinese labs proved it's possible to do on a small budget, efficiently.

Why is me using GPL snippets in my code and further distributing my code as GPL ok, but using GPL snippets in training an LLM and further distributing that LLM as GPL wrong?

As I said above, it seems it's an intellectually dishonest argument

@budududuroiu @tante If you use GPL snippets and comply with the GPL by publishing the work you incorporated it into, you’ve honored the agreement that gave you those snippets in the first place. Good.

An LLM scraping them all, mashing them together, and regurgitating it out of context does not honor the gift of the code. It’s also a violation of the GPL, which uses copyright to enforce the continued freedom of that code.

GPL licensing the model doesn’t help.

@DataGhost @tante

> It’s also a violation of the GPL

Under what? GPL gives you the freedom to run, study, share, or modify the software as long as you publish the derivative work as GPL too. Releasing a truly open source model with training data, training scripts, hyperparameters, and everything needed to reproduce the model would not violate GPL.

@budududuroiu @tante The output of that open model will incorporate the GPL code it was trained on, generate derivative works, and strip all credit and licensing.

Apparently this is an unsettled area of copyright law, so please set aside “violation of the GPL”.

https://www.gnu.org/licenses/gpl-faq.en.html#:~:text=video.-,The%20exception%20would%20be%20when%20the%20program%20displays%20a%20full%20screen%20of%20text%20and/or%20art%20that%20comes%20from%20the%20program.,-Then

Frequently Asked Questions about the GNU Licenses - GNU Project - Free Software Foundation

@DataGhost @tante

> generate derivative works, and strip all credit and licensing.

Isn't this already an issue we solve by expecting developers to not strip licensing and credit?

Also, GPL-ness isn't static, when GPL didn't prevent projects from offering managed solutions for GPL code as SaaS, people came up with AGPL.

@budududuroiu @tante Do we expect LLMs to not strip licensing and credit?

It seems fundamental to their operation.

@DataGhost @tante it is also fundamental to their operation that they get prompted by a user, LLMs don't have initiative, they respond to prompts from users that can be held accountable

@tante Yeah, if anything it'll do the opposite. Lowering entry costs will drive people out of the market, so that the only ones who can afford to get stuff actually made by a human are the ones who have acquired sufficient money / power to afford it.

E.g. Once upon a time, every chair or table would have been hand-made. Where does hand-made furniture sit in the market, now?

@tante Anyone in tech who tosses around the term 'Democratising' means they want to remove the need for consent, the ability for creative people to refuse to allow their art be used to put a glossy sheen on the indefensible. It's the same language we hear from incels when it comes to sex, what their involuntary partner may want isn't even a consideration.
@tante Indeed, it's the opposite of democratization. The reason: Who owns the models, who owns the data centers? Oligarchs have control over whether the models generate and what they generate.
This is spot on. AI will automate the boring parts but the "democratization" crowd always conflates access with skill. The tools get better, the bar gets higher, not lower

@tante

When we look at any piece of technology being offered to us the very first thing we should ask is who ends up in control of the process / who is made dependent on the process?

Convenience can only be considered beneficial when we can trust the service not to screw us

@tante exactly the opposite of democratizing https://mastodon.social/@noleli/116104276740757370
@noleli @tante It was ever thus with big tech's claim of "democratizing" anything
@thrilway @tante indeed, ever since they rejected the promise of the personal computer revolution and started clawing it back with “cloud services”
@tante Universal Basic Income

@tante Democratize? This means bypassing actual experience, exploiting the experience of those who did put in the time?

I hate these tech bro terms thst attempt to make their awful things seem cool and innovative rather than exploitive.

@tante speaking as a parent with questionable time management skills and a creative passion I can 10000% verify this
@tante the "AI democratisation" narrative is also just toxic as fuck because it always portrays people with certain skills as an evil cabal gatekeeping those skills, turning people on each other to not focus on the real issues.
@tante renting a skill instead of learning it is the opposite of empowering
@tante They said the same thing about crypto - remember the "decentralised web"?? Anyone? Anyone?
@tante give people UBI and like half of such problems are solved immediately
@tante The Bros are giving their auto-complete-bots time for creativity...
@crenquis yeah I saw that on bsky. These guys are super not okay

@tante

Just like the cotton factories "democratized" weaving.

Anybody could do it.

Of course, they had to do it 12 hours a day, destroy their lungs and hearing, for a tenth of the pay.

@Dianora