cross-posted from: https://lemmy.blahaj.zone/post/45981744
Breaking news: mere hours ago (stretching the definition of breaking, I know) the new UK Prime Penister — ahem! Minister — announced new classes for children age 14 including classes on AI.
Now, to me this could go either way. Either, it opens these children’s eyes to the problems with AI and actually encourages the next generation to solve the issues of AI, or this lazy generation might end up being indoctrinated into the greed and corruption surrounding AI. Knowing the fallibility of human beings, in my humble opinion it is most certainly the latter, but who knows? What are your thoughts on this announcement?
We should make it mandatory that in order to legislate something like this you have to prove you actually understand how it works
I would love to hear this prime minister describe to me what “AI” can and can’t do in his own words.
And that is exactly what I call hitting the nail on the head. I completely agree with you. Imagine any politician try to explain AI
“So Prime Minister, Can you explain how AI works?”
“Well, I, er… I’m not going to comment on policies I’m pushing but… this afternoon we nationalised pretzels…” 🙄
My kingdom for a journalist with some chutzpah to actually ask these questions too
Half these idiots dont even realize the “AI” theyre plugging is basically predictive text/syncophancy machines, the REAL AI that matters that helps scientists with shit like datasets isnt coming from Meta.
Secondary school pupils in England will be able to study subjects like manufacturing and AI in addition to core academic learning under reforms outlined by Prime Minister Andy Burnham.
Huh.
considers
I’m still a little dubious that we’re at the point that it makes a lot of sense to dump a lot of people into AI applications. Like, yes, please put as many people as possible on the core research, the kind of thing that people at, say, Nvidia or Google are doing. Figure out how to train a model more-efficiently — that’d be great. Fundamental work towards AGI? Please. New data structures for machine learning other than multilayer neural nets? That’d be great, all important research areas. But “writing programs that make use of existing AI systems and utilizing existing AI systems day-to-day”…the problem I have is that this is early, and the thing is in a state of rapid change. We’re still knocking out low-hanging fruit, solving problems. You build a workflow that uses some model that exists today, and even if the model continues to be available, the limitations and workarounds for a model twelve months later may be quite different. I’d want that to stabilize before shifting a lot of people into using models, so they aren’t having to re-learn a lot of stuff.
Like, Stable Diffusion 1.5 was released in late 2022. The problems that you’d see with that model and the prompting techniques to deal with it and work around limitations are very much not the same problems and limitations that exist for Flux 2 in 2026. Stable Diffusion supports token weighting; Flux does not. Stable Diffusion 1.5 performed poorly with English sentences, wanted a list of tokens; Flux does not. Stable Diffusion 1.5 had problems with fingers; Flux 2 does not. If you spend time getting someone to learn how to operate a diffusion model of 2026, is that skillset going to be similarly relatively-obsolete in another three years? Because if so, the student has to be providing an enormously high return-on-educational-investment over each of those three years to make it sensible to invest lots of time in making people expert in using that model.
I’m not saying that nobody should learn to use them, but I don’t know if this is a “mobilize government and shift things ASAP across society” situation. And that’s not because I don’t think that AI is important. It’s just because I think that from a usage standpoint, things are changing very quickly right now.
The best skills to teach someone, all else held equal, are the skills that they can use for their entire working life. I strongly suspect that the great bulk of LLM or diffusion model user skillset is going to be quite different in a relatively-short period of time, and I can very readily believe that we won’t even be using the relatively-primitive “talk to a raw LLM” within a decade or two, have some considerably-more-sophisticated system moving towards AGI.
Yes, pretty much any skill has value. But lets say that you have the time and funds to teach a student, oh…statistics or how to use an LLM. Which is going to provide the best return on investment for the educational time? Because that’s the real metric that matters.
And especially with the way schools are in England right now (cut funding, staff shortages, institutional instability, rising delinquent behaviour) schools need to prioritise what’s most important to teach now more than ever if you ask me.
To put even more context on that…when I first used a diffusion model, I was using the Automatic1111 frontend (well, okay, actually, I think it was a variant, but the first one I put much time into). I was using Stable Diffusion 1.5 or a derived model, can’t recall.
Automatic1111 looks more-or-less like what I think most of the cloud service systems look like. At its core, you plug in a text prompt, you click a button, you get an image.
In 2026, I pretty much only use ComfyUI, which has a directed-acyclic-graph-based system akin to what some image processing software or visual-programming systems use. IIRC, GIMP uses a DAG internally these days, though doesn’t directly expose it to users. searches Yeah, GEGL. And AFAICT, that’s generally where the momentum is — it’s not just me. That’s a lot more powerful for making changes early in a workflow and re-running it, though more difficult to learn. And from a user standpoint, it looks wildly different.
I’m not saying that there was no transferable knowledge there. You’re still working with a prompt somewhere in there. And the parameters that go to models are similar. But…you’re talking about what is effectively a new syntax for prompts. A wildly different set of limitations to work around with those prompts. Shifting from a UI that is basically an “input a string and click go” to a visual programming language.
And…I haven’t even listed some of the other of the changes. Uh, let’s see. Stable Diffusion 1.5 supports negative prompts; one can list prompt terms that should not be presented in an image; Flux does not. Flux has some sort of natural-language-processing black magic that frequently lets one use English-language descriptions to do things like “Divide an image into five separated panels. The first panel should contain…”. With SD 1.5, one would have to use some sort of plugin to subdivide an image and manually apply different prompts to different regions of an image.
Like, virtually the only thing in common between the two are sampler settings (and I use different sampler settings with each model) and the fact that both have a prompt, and there’s virtually no overlap in prompt syntax. And that was over a timespan of three years. Well, less than three years.
I can believe that you can teach someone to directly use a model. I can believe that you can teach someone to write software that uses a model. I can believe that they can produce useful output using that model. But where I have some very real problems is buying into the idea that in 2026, we’ve reached the point where the user skillset is going to be long-lasting enough that this is where one is going to get the best return. It’s like…I don’t know, training people to fly the Wright Flyer en masse:
https://en.wikipedia.org/wiki/Wright_Flyer
The Flyer design depended on wing-warping controlled by a hip cradle under the pilot, and a foreplane or “canard” for pitch control, features which would not scale and produced a hard-to-control aircraft.
You’d be right if you predicted in 1905 that heavier-than-air aircraft were going to be a very important thing in the future, but it was extremely early days then, and the control interface and the challenges that the pilot faces are probably going to be very different in short order.
Why would you learn about it when it’s poised to replace you?
