LBeaudoux
Armin RonacherThe future is unevenly distributed.
dtanzer@mitsuhiko Interesting. But not totally unexpected. People here seem to be more skeptical, which is not necessarily a bad thing.
@dtanzer It's also not necessarily a good thing. There is a reason people over here are called luddites at times.
@mitsuhiko I didn't say it's good 🙂
I personally don't understand people who do not want to try those tools, but I understand those who try and don't use for a reason.
I, personally, am torn between "wow, this is cool" and "damn, this needs too much babysitting" (ignoring all the society-level, etc. effects for the sake of this conversation) Use occasionally.
Eugen Rochko@mitsuhiko @dtanzer You’re saying that like it’s a bad thing 🙂
@Gargron I can't speak for @mitsuhiko of course, but for me it's not a bad thing, but also not necessarily a good thing (again - see above 😉 ).
But in this case, I think there is not a big case against waiting for the hype to calm down: Those tools are very easy to learn, and *everyone* is winging it right now anyway.
@dtanzer @Gargron I happen to think that there is very strong case to make that it's a bad idea to wait for the hype to calm down, particularly for politicians and decision makers. Late movers will be punished in this space. Particularly for us Europeans we will become hyper dependent on other countries and it will become close to impossible for us to compete over time, particularly if the foundation labs decide to close the door behind themselves through the legal systems.
@mitsuhiko
I was not talking about politicians and decision makers. In your original post, you asked whether we were personally using it...
@mitsuhiko Here's something I wrote earlier in the thread: "I personally don't understand people who do not want to try those tools, but I understand those who try and [then] don't use for a reason."
Goseg Train Thread@mitsuhiko @dtanzer @Gargron AI proponents say it's so easy to master that even a plumber can do it, and also that we all better hurry up and become familiar with it now or we'll get left behind. It's gotta be one or the other, can't be both.
@wvistheduck @dtanzer @Gargron maybe some do, but my messaging on this has been that it takes time and effort. It’s not as straightforward as some people say.
privilegejunkie@mitsuhiko @wvistheduck @dtanzer @Gargron Compared to obtaining a computer science degree and learning system design, learning AI tooling is very easy. I’ve never felt the panic of falling behind in this space to be real at an individual contributor level.
IBBoard@mitsuhiko @dtanzer @Gargron Politicians and policy makers absolutely shouldn't be late movers.
They need to get on top of things and legislature appropriately before the shit hits the fan and the LLM companies shrug and say either "we told you we didn't guarantee anything" or "externalities like society and the environment aren't our problem when we're making investors money in the short term".
That's not the same as "they need to listen to the hype from the companies selling it and throw all of their support and national investment behind it with a blank cheque to do what they want".
@mitsuhiko @dtanzer @Gargron Yes, because "do what you want and screw the societal side-effects" approaches from governments have worked out so well for the world in the past when applied to companies that are 100% in it for the profit and not concerned in the slightest about the collateral damage along the way 😐
Remind me how well that went with the oil industry, the tobacco industry, leaded petrol, social media at large, cryptocurrency and NFTs, right-wing propaganda-as-news, …
@mitsuhiko @dtanzer @Gargron By all means, encourage European governments to invest in machine learning, robotics and other valuable areas. And encourage digital sovereignty rather than relying on a small group of US megacorps.
But that's not the same as "throw money at LLMs because they're the current hot thing". Bitcoin and NFTs were also supposed to be the big thing at one point. And luckily most governments didn't throw everything at those.
@mitsuhiko @ibboard @dtanzer This technology has yet to prove it has any benefits (beyond the "vibes" of the people pushing it). So far it's been forcefully added into most products against the wishes of the users, to prop up a speculative stock market--that's not a sign that it's a desirable technology. And we are only beginning to see the damage it's doing through skill atrophy--that's going to be a long term effect. So no, I don't think we need investments.
@Gargron @ibboard @dtanzer That's the weirdness of it. All the consumer AI slop is completely off-putting, but what coding agents are capable of is on a completely different level. It's going to move labor massively towards capital sadly and more worrying, it's going to move all the gains to the US and China and away from Europe. No machine will pay taxes here.
Paul Moore@mitsuhiko @Gargron @ibboard @dtanzer But aren't the people claiming "good" AI results all using expensive subscription versions? Don't the "AI is good" claims all ultimately boil down to an elitist position of "programming is only accessible to those who can afford it"?
Chris L. Barnes@pfmoore @mitsuhiko @Gargron @ibboard @dtanzer Yes, in the same way that X years ago it was most accessible to people with reliable internet connections, and Y years ago it was most accessible to people with machines powerful enough to compile locally rather than sending punchcards in the mail, and Z years ago it was people who had access to particular elite education. Being on the cutting edge of technology has never been equitable.
@mitsuhiko @Gargron @ibboard @dtanzer So stop saying that AI can provide good results, then? Doing that simply promotes the interests of those large US corporations. Instead, point out that AI isn't ready for common acceptance because it's gated behind the ability to pay exorbitant subscriptions. Don't subsidise AI companies, demand that AI tools are in line with existing coding tools (i.e., free or open source).
Anil Dash@pfmoore @mitsuhiko @Gargron @ibboard @dtanzer you’re having conversations about 2 separate spheres of economic concern. You may be right about individual-level cost. But the nation-state-level investment needed to create sufficient infrastructure to avoid a colonialist relationship with either the U.S. or China isn’t present yet, and both are trying to create that dependency, quickly. That’s the thing to hedge against.
@pfmoore @mitsuhiko @Gargron @ibboard @dtanzer you’re having conversations about 2 separate spheres of economic concern. You may be right about individual-level cost. But the nation-state-level investment needed to create sufficient infrastructure to avoid a colonialist relationship with either the U.S. or China isn’t present yet, and both are trying to create that dependency, quickly. That’s the thing to hedge against.
@anildash @pfmoore @mitsuhiko @Gargron @dtanzer But investing huge amounts (which society can't afford) into "ChatGPT but European" (which we arguably won't get them what people say it will) isn't the only way to avoid that dependency.
It's only a colonialist dependency if you become dependent on it. If people push back on "GenAI crowbarred into everything (regardless of need)", if governments block GenAI companies until they prove they're dealing with biases and harms, and if people examine "what it can actually do" versus "what people think it can do because of hype and eloquence" (and what's solid, probably ML versus hyped "AI") then we have a different way out.
If governments invested in everything that was touted as the next big thing that we can't let the USA/China have then we'd still be paying the bill for cryptocurrency investment (amongst other things).
"Don't miss out on the hype train!" isn't a way to run a government. "Make sure the hype train follows the rules while working out whether it is actually going somewhere or whether it's at its limit already (and demonstrably not doing a lot of what people think it is)" is.
@pfmoore @mitsuhiko @Gargron @ibboard @dtanzer you’re having conversations about 2 separate spheres of economic concern. You may be right about individual-level cost. But the nation-state-level investment needed to create sufficient infrastructure to avoid a colonialist relationship with either the U.S. or China isn’t present yet, and both are trying to create that dependency, quickly. That’s the thing to hedge against.
@pfmoore @mitsuhiko @Gargron @ibboard @dtanzer you’re having conversations about 2 separate spheres of economic concern. You may be right about individual-level cost. But the nation-state-level investment needed to create sufficient infrastructure to avoid a colonialist relationship with either the U.S. or China isn’t present yet, and both are trying to create that dependency, quickly. That’s the thing to hedge against.
yoasif@mitsuhiko @pfmoore @Gargron @ibboard @dtanzer So you are already paying the LLM pirates, and your livelihood depends on piracy, so your proposal is simply to make the models cheaper for you.
It is a foregone conclusion that the tools are good, and legislation is meaningless, except as a cudgel against foreign competition.
@Gargron @mitsuhiko @pfmoore @ibboard @dtanzer it’s not at all uncommon these days. There are basically two cohorts now, those who create by coding, and those who create by orchestrating systems that output code. If it helps, for them, it’s a bit like having interpreted languages where 99% of users have no idea what the actual assembly code looks like. Paying for that abstraction provides value to these users.
Joe@hungryjoe @Gargron @anildash @mitsuhiko @pfmoore @dtanzer None of the companies are profitable… YET.
If we all just keep using it unquestioningly and throwing money at it and accepting price increases while ignoring societal harm then MAYBE we can start making money for a bunch of investors so that they can afford to insulate themselves from the societal and environmental harm that it causes 👍
@Gargron @mitsuhiko @pfmoore @ibboard @dtanzer it’s not at all uncommon these days. There are basically two cohorts now, those who create by coding, and those who create by orchestrating systems that output code. If it helps, for them, it’s a bit like having interpreted languages where 99% of users have no idea what the actual assembly code looks like. Paying for that abstraction provides value to these users.
Jonathan Hendry@mitsuhiko @Gargron @ibboard @dtanzer
Machines wouldn't pay taxes if the EU invested, either.
@mitsuhiko @Gargron @ibboard @dtanzer
Do they though.
@mitsuhiko @ibboard @dtanzer @Gargron
I suppose they should have jumped on blockchain and NFTs, and the Metaverse, too. Massive waste of taxpayer money and huge transfer of wealth to sleazy venture capitalists.
@mitsuhiko @ibboard @dtanzer @Gargron
Sam Altman was one of the people promoting crypto blockchain nonsense, specifically "worldcoin".
The VC firms pushing crypto and NFTs were the same ones pushing AI.
Thomas Sturm@mitsuhiko @dtanzer @Gargron You will only be hyper dependent on AI if you forget how to code. I write my code and know what it does.
I’m not going to have to explain to any boss or client ever that I can’t debug my code because some AI is having a bad day.
Andrei Kucharavy1. It was not exactly open to start with. To get the foundation of any coding LLM you needed to rip off tons of repos, Q&As, and blogs that were all protected IP. Which flies only if you are in the US and have enough money for your lawyers to fight god and your lobbyists to buy the entirety of Congress and Senate, or are in a country where IP legislation is inexistent or not enforced (China, Russia, …).
2. I my experience, the main problem of Mastodon users is not with the tech, but with externalities: the ethics of how it is built, to whom the money from its goes, and how this affects the understandability and maintainability of the body of code we all rely on.
Which makes sense - you are talking to a social media made almost exclusively of people who said fuck it to other social media for the same externality reasons.
In my experience FOSS compliance-first LLMs here receive a massive positive feedback, while public on most other social media does not care about it and is driven by pretty much exclusively by shiny feature demos.
This ethics vs shine split is almost universal across topics on Fedi vs elsewhere.
Stefan Bohacek@Gargron @mitsuhiko @dtanzer I thought it's become a common knowledge that Luddites were a workers' rights movement, not an anti-tech one.
https://www.smithsonianmag.com/history/what-the-luddites-really-fought-against-264412/
(((H. N.)))@mitsuhiko @dtanzer You’re saying that like Luddites were wrong, which is not the history I was taught.
Konosocio@mitsuhiko @dtanzer historically speaking, the luddites were right.
@konosocio @dtanzer Historically speaking, luddites had to come to terms with the fact that it was entirely irrelevant if they were right or wrong, automation was the future that the world demanded.
@mitsuhiko @dtanzer that's one way to look at it. Alternatively, they were defeated by a powerful coalition of state and capital which led directly to today's unprecedented inequality between the owner and the working class. And that means that we are still fighting the rightous fight of the luddites.
Quantum ⊂ AI@mitsuhiko @dtanzer you know the Luddites were Correct, right?
bjb 
Luddites were a social/labour movement, not an anti-tech movement.
Anyway, on this topic, I listened to the latest episode of an AI podcast, and the hosts agreed that LLMs are not going to get to the point where they can solve problems, and they can be the human language front end for machine learning systems that will solve all your problems for you. BTW, those are specialized models and need to be customized for the type of task.
So thousands of workers ...
1/n
... in lots of companies wasted their time being frustrated with half- (or less) baked tech that caused problems, for the last two + years, because these LLMs were being foisted on us when we could see it wasn't going to work. What a huge waste of resources, environmental and individual corporate resources.
We could have been using tools that give the same answer to the same question, efficiently, without spouting manipulative stomach-churning brown-nosing language.
Ricardo Carvalho@mitsuhiko I’d totally expected it, mastodon people are more purists 😛
I am also not impressed on people here being kind-of scared of people they regarded as purists and gods of code to also embrace AI (Linus Torvalds and antirez for example)
Bite Code!@mitsuhiko On one hand, there is a lot of identity and value signaling attached to not use AI in these circles. Big "holier than thou." culture.
But also, on the positive side, ethics are important to this community, and it's true AI causes massive ethics issues. I can understand not using AI because you don't want to funnel money and attention to a system that you deem unethical.
As a vegetarian, I understand the struggle. Striking the balance between integration and values is hard.
Tfe 
@mitsuhiko Eta nola zenbatu behar dugu "Tab" bat erabiltzea 10 lerro idazteko? Nire kodigoa al da edo AA-rena?
Adrian Cockcroft@mitsuhiko Thanks for running this comparison. LinkedIn is similar to X as far as I can see. Also I just had a demo of a new high quality product with masses of functionality that was built in the last few months by one (very experienced CTO) guy and teams of carefully managed agents. It can be done right, even if most people+agents are generating slop.
Alrecenk@mitsuhiko This reminds me of a saying I've heard in game development circles: There are a lot of game developers, but very few people who actually develop games. I imagine it's the same for software development broadly.
@Alrecenk From where I'm sitting, anyone who uses an agent to ship stuff is a developer. Maybe not a particularly good one, but a developer nonetheless. If you paint, even if you paint badly, you are still a painter. No gatekeeping there
@mitsuhiko I'm not talking about people who ship. I'm talking about people who talk a lot and never ship anything.