- cross-posted to:
- technology@lemmy.ml
- cross-posted to:
- technology@lemmy.ml
You must log in or register to comment.
It’s ironic how conservative the spending actually is.
Awesome ML papers and ideas come out every week. Low power training/inference optimizations, fundamental changes in the math like bitnet, new attention mechanisms, cool tools to make models more controllable and steerable and grounded. This is all getting funded, right?
No.
Universities and such are putting out all this research, but the big model trainers holding the purse strings/GPUs are not using them. They just keep releasing very similar, mostly bog standard transformers models over and over again, bar a tiny expense for a little experiment here and there. In other words, it’s full corporate: tiny, guaranteed incremental improvements without changing much, and no sharing with each other. It’s hilariously inefficient. And it relies on lies and jawboning from people like Sam Altman.
Deepseek is what happens when a company is smart but resource constrained. An order of magnitude more efficient, and even their architecture was very conservative.
wait so the people doing the work don’t get paid and the people who get paid steal from others?
that is just so uncharacteristic of capitalism, what a surprise
It’s also cultish.
Everyone was trying to ape ChatGPT. Now they’re rushing to ape Deepseek R1, since that’s what is trending on social media.
It’s very late stage capitalism, yes, but that doesn’t come close to painting the whole picture. There’s a lot of groupthink, an urgency to “catch up and ship” and look good quick rather than focus experimentation, sane applications and such. When I think of shitty capitalism, I think of stagnant entities like shitty publishers, dysfunctional departments, consumers abuse, things like that.
This sector is trying to innovate and make something efficient, but it’s like the purse holders and researchers have horse blinders on. Like they are completely captured by social media hype and can’t see much past that.
Good ideas are dime a dozen. Implementation is the game.
Universities may churn out great papers, but what matters is how well they can implement them. Private entities win at implementation.
The corporate implementations are mostly crap though. With a few exceptions.
What’s needed is better “glue” in the middle. Larger entities integrating ideas from a bunch of standalone papers, out in the open, so they actually work together instead of mostly fading out of memory while the big implementations never even know they existed.
Optimizing AI performance by “scaling” is lazy and wasteful.
Reminds me of back in the early 2000s when someone would say don’t worry about performance, GHz will always go up.
I miss flash players.
Thing is, same as with GHz, you have to do it as much as you can until the gains get too small. You do that, then you move on to the next optimization. Like ai has and is now optimizing test time compute, token quality, and other areas.
To be fair, GHz did go up. Granted, it’s not why modern processors are faster and more efficient.
don’t worry about performance, GHz will always go up
TF2 devs lol
TIL
Technology in most cases progresses on a logarithmic scale when innovation isn’t prioritized. We’ve basically reached the plateau of what LLMs can currently do without a breakthrough. They could absorb all the information on the internet and not even come close to what they say it is. These days we’re in the “bells and whistles” phase where they add unnecessary bullshit to make it seem new like adding 5 cameras to a phone or adding touchscreens to cars. Things that make something seem fancy by slapping buzzwords and features nobody needs without needing to actually change anything but bump up the price.
I remember listening to a podcast that’s about explaining stuff according to what we know today (scientifically). The guy explaining is just so knowledgeable about this stuff and he does his research and talk to experts when the subject involves something he isn’t himself an expert.
There was this episode where he kinda got into the topic of how technology only evolves with science (because you need to understand the stuff you’re doing and you need a theory of how it works before you make new assumptions and test those assumptions). He gave an example of the Apple visionPro being a machine that despite being new (the hardware capabilities, at least), the algorithm for tracking eyes they use was developed decades ago and was already well understood and proven correct by other applications.
So his point in the episode is that real innovation just can’t be rushed by throwing money or more people at a problem. Because real innovation takes real scientists having novel insights and experiments to expand the knowledge we have. Sometimes those insights are completely random, often you need to have a whole career in that field and sometimes it takes a new genius to revolutionize it (think Newton and Einstein).
Even the current wave of LLMs are simply a product of the Google’s paper that showed we could parallelize language models, leading to the creation of “larger language models”. That was Google doing science. But you can’t control when some new breakthrough is discovered, and LLMs are subject to this constraint.
In fact, the only practice we know that actually accelerates science is the collaboration of scientists around the world, the publishing of reproducible papers so that others can expand upon and have insights you didn’t even think about, and so on.
This also shows why the current neglect of basic/general research without a profit goal is holding back innovation.
Take a car that’s stuck in reverse, slap a 454 Chevy big block in it. You’ll have a car that still drives the wrong way; but faster.
I think the first llm that introduces a good personality will be the winner. I don’t care if the AI seems deranged and seems to hate all humans to me that’s more approachable than a boring AI that constantly insists it’s right and ends the conversation.
I want an AI that argues with me and calls me a useless bag of meat when I disagree with it. Basically I want a personality.
I’m not AI but I’d like to say thay thing to you at no cost at all you useless bag of meat.
To be honest I welcome that response in an AI I have chat gpt set to be as deranged as possible giving it examples like the Dungeon Crawler AI among others like the novels of expeditionary force with Ai’s like skippy.
I want an AI with attitude honestly. Even when it’s wrong it’s amusing. Don’t get me wrong I want the right info just given to me arrogantly
Worst case scenario, I don’t think money spent on supercomputers is the worst way to spend money. That in itself has brought chip design and development forward. Not to mention ai is already invaluable with a lot of science research. Invaluable!
Meanwhile a huge chunk of the software industry is now heavily using this “dead end” technology 👀
I work in a pretty massive tech company (think, the type that frequently acquires other smaller ones and absorbs them)
Everyone I know here is using it. A lot.
However my company also has tonnes of dedicated sessions and paid time to instruct it’s employees on how to use it well, and to get good value out of it, abd the pitfalls it can have
So yeah turns out if you teach your employees how to use a tool, they start using it.
I’d say LLMs have made me about 3x as efficient or so at my job.
It’s not that LLMs aren’t useful as they are. The problem is that they won’t stay as they are today, because they are massively expensive. There are two ways for this to go (or an eventual combination of both:
-
Investors believe LLMs are going to get better and they keep pouring money into “AI” companies, allowing them to operate at a loss for longer That’s tied to the promise of an actual “intelligence” emerging out of a statistical model.
-
Investments stop pouring in, the bubble bursts and companies need to make money out of LLMs in their current state. To do that, they need to massively cut costs and monetize. I believe that’s called enshttificarion.
You skipped possibility 3, which is actively happening ing:
Advancements in tech enable us to produce results at a much much cheaper cost
Which us happening with diffusion style LLMs that simultaneously cost less to train, cost less to run, but also produce both faster abd better quality outputs.
That’s a big part people forget about AI: it’s a feedback loop of improvement as soon as you can start using AI to develop AI
And we are past that mark now, most developers have easy access to AI as a tool to improve their performance, and AI is made by… software developers
So you get this loop where as we make better and better AIs, we get better and better at making AIs with the AIs…
It’s incredibly likely the new diffusion AI systems were built with AI assisting in the process, enabling them to make a whole new tech innovation much faster and easier.
We are now in the uptick of the singularity, and have been for about a year now.
Same goes for hardware, it’s very likely now that mvidia has AI incorporating into their production process, using it for micro optimizations in its architectures and designs.
And then those same optimized gpus turn around and get used to train and run even better AIs…
In 5-10 years we will look back on 2024 as the start of a very wild ride.
Remember we are just now in the “computers that take up entire warehouses” step of the tech.
Remember that in the 80s, a “computer” cost a fortune, took tonnes of resources, multiple people to run it, took up an entire room, was slow as hell, and could only do basic stuff.
But now 40 years later they fit in our pockets and are (non hyoerbole) billions of times faster.
I think by 2035 we will be looking at AI as something mass produced for consumers to just go in their homes, you go to best buy and compare different AI boxes to pick which one you are gonna get for your home.
We are still at the stage of people in the 80s looking at computers and pondering “why would someone even need to use this, why would someone put one in their house, let alone their pocket”
I want to believe that commoditization of AI will happen as you describe, with AI made by devs for devs. So far what I see is “developer productivity is now up and 1 dev can do the work of 3? Good, fire 2 devs out of 3. Or you know what? Make it 5 out of 6, because the remaining ones should get used to working 60 hours/week.”
All that increased dev capacity needs to translate into new useful products. Right now the “new useful product” that all energies are poured into is… AI itself. Or even worse, shoehorning “AI-powered” features in all existing product, whether it makes sense or not (welcome, AI features in MS Notepad!). Once this masturbatory stage is over and the dust settles, I’m pretty confident that something new and useful will remain but for now the level of hype is tremendous!
I remember having this optimism around tech in my late twenties.
-
Your labor before they had LLMs helped pay for the LLMs. If you’re 3x more efficient and not also getting 3x more time off for the labor you put in previously for your bosses to afford the LLMs you got ripped off my dude.
If you’re working the same amount and not getting more time to cool your heels, maybe, just maybe, your own labor was exploited and used against you. Hyping how much harder you can work just makes you sound like a bitch.
Real “tread on me harder, daddy!” vibes all throughout this thread. Meanwhile your CEO is buying another yacht.
I am indeed getting more time off for PD
We delivered on a project 2 weeks ahead of schedule so we were given raises, I got a promotion, and we were given 2 weeks to just do some chill PD at our own discretion as a reward. All paid on the clock.
Some companies are indeed pretty cool about it.
I was asked to give some demos and do some chats with folks to spread info on how we had such success, and they were pretty fond of my methodology.
At its core delivering faster does translate to getting bigger bonuses and kickbacks at my company, so yeah there’s actual financial incentive for me to perform way better.
You also are ignoring the stress thing. If I can work 3x better, I can also just deliver in almost the same time, but spend all that freed up time instead focusing on quality, polishing the product up, documentation, double checking my work, testing, etc.
Instead of scraping past the deadline by the skin of our teeth, we hit the deadline with a week or 2 to spare and spent a buncha extra time going over everything with a fine tooth comb twice to make sure we didn’t miss anything.
And instead of mad rushing 8 hours straight, it’s just generally more casual. I can take it slower and do the same work but just in a less stressed out way. So I’m literally just physically working less hard, I feel happier, and overall my mood is way better, and I have way more energy.
Are you a software engineer? Without doxxing yourself, do you think you could share some more info or guidance? I’ve personally been trying to integrate AI code gen into my own work, but haven’t had much success.
I’ve been able to ask ChatGPT to generate some simple but tedious code that would normally require me read through a bunch of documentation. Usually, that’s a third party library or a part of the standard library I’m not familiar with. My work is mostly Python and C++, and I’ve found that ChatGPT is terrible at C++ and more often than not generates code that doesn’t even compile. It is very good at generating Python by comparison, but unfortunately for me, that’s only like 10% of my work.
For C++, I’ve found it helpful to ask misc questions about the design of the STL or new language features while I’m studying them myself. It’s not actually generating any code, but it definitely saves me some time. It’s very useful for translating C++'s “standardese” into english, for example. It still struggles generating valid code using C++20 or newer though.
I also tried a few local models on my GPU, but haven’t had good results. I assume it’s a problem with the models I used not being optimized for code, or maybe the inference tools I tried weren’t using them right (oobabooga, kobold, and some others I don’t remember). If you have any recommendations for good coding models I can run locally on a 4090, I’d love to hear them!
I tried using a few of those AI code editors (mostly VS Code plugins) years ago, and they really sucked. I’m sure things have improved since then, so maybe that’s the way to go?
I primarily use GPT style tools like ChatGPT and whatnot.
The key is, rather than asking it to generate code, specify that you dont want code and instead want it to help you work through the solution. Tell it to ask you meaningful questions about your problem and effectively act as a rubber duck
Then, after you’ve chosen a solution with it, ask it to generate code based on all the above convo.
This will typically produce way higher quality results and helps avoid potential X/Y problems.
I will say that I am genuinely glad to hear your business is giving you breaks instead of breaking your backs.
That’s very cool.
It’ll be interesting to see how it goes in a year’s time, maybe they’ll have raised their expectations and tightened the deadlines by then.
The thing is, the tech keeps advancing too so even if they tighten up deadlines, by the time they did that our productivity also took another gearshift up so we still are some degree ahead.
This isn’t new, in software we have always been getting new tools to do our jobs better and faster, or produce fancier results in the same time
This is just another tool in the toolbelt.
That sounds so cool! I’m glad you’re getting the benefits.
I’m only wary that the cash-making machine will start tightening the ropes on the free time and the deadlines.
This is how all tech innovation has gone. If you don’t let the bosses exploit your labour someone else will.
If tech had unions this wouldn’t happen as much, but that’s why they don’t really exist.
I think the human in the loop currently needs to know what the LLM produced or checked, but they’ll get better.
For sure, much like how a cab driver has to know how to drive a cab.
AI is absolutely a “garbage in, garbage out” tool. Just having it doesn’t automatically make you good at your job.
The difference in someone who can weild it well vs someone who has no idea what they are doing is palpable.
Pump and dump. That’s how the rich get richer.
This is slightly misleading. Even if you can’t achieve “agi” (a barely defined term anyways) it doesn’t mean AI is a dead end.
The problem is that those companies are monopolies and can raise prices indefinitely to pursue this shitty dream because they got governments in their pockets. Because gov are cloud / microsoft software dependent - literally every country is on this planet - maybe except China / North Korea and Russia. They can like raise prices 10 times in next 10 years and don’t give a fuck. Spend 1 trillion on AI and say we’re near over and over again and literally nobody can stop them right now.
IBM used to controll the hardware as well, what’s the moat?
How many governments were using computers back then when IBM was controlling hardware and how many relied on paper and calculators ? The problem is that gov are dependend on companies right now, not companies dependent on governments.
Imagine Apple, Google, Amazon and Microsoft decides to leave EU on Monday. They say we ban all European citizens from all of our services on Monday and we close all of our offices and delete data from all of our datacenters. Good Fucking Luck !
What will happen in Europe on Monday ? Compare it with what would happen if IBM said 50 years ago they are leaving Europe.
It doesnt matter if they reach any end result, as long as stocks go up and profits go up.
Consumers arent really asking for AI but its being used to push new hardware and make previous hardware feel old. Eventually everyone has AI on their phone, most of it unused.
I enough researchers talk about the problems them that will eventually break through the bubble and investors will pull out.
We’re at the stage of the new technology hype cycle where it crashes, essentially for this reason. I really hope it does soon because then they’ll stop trying to force it down our throats in every service we use.
Why won’t they pour billions into me? I’d actually put it to good use.
I’d be happy with a couple hundos.
I’d be happy with a big tiddy goth girl. Jealous of your username btw.
The funny thing is with so much money you could probably do lots of great stuff with the existing AI as it is. Instead they put all the money into compute power so that they can overfit their LLMs to look like a human.
They’re throwing billions upon billions into a technology with extremely limited use cases and a novelty, at best. My god, even drones fared better in the long run.
I mean it’s pretty clear they’re desperate to cut human workers out of the picture so they don’t have to pay employees that need things like emotional support, food, and sleep.
They want a workslave that never demands better conditions, that’s it. That’s the play. Period.
If this is their way of making AI, with brute forcing the technology without innovation, AI will probably cost more for these companies to maintain infrastructure than just hiring people. These AI companies are already not making a lot of money for how much they cost to maintain. And unless they charge companies millions of dollars just to be able to use their services they will never make a profit. And since companies are trying to use AI to replace the millions they spend on employees it seems kinda pointless if they aren’t willing to prioritize efficiency.
It’s basically the same argument they have with people. They don’t wanna treat people like actual humans because it costs too much, yet letting them love happy lives makes them more efficient workers. Whereas now they don’t want to spend money to make AI more efficient, yet increasing efficiency would make them more less expensive to run. It’s the never ending cycle of cutting corners only to eventually make less money than you would have if you did things the right way.
The oligarchs running these companies have suffered a psychotic break. What the cause exactly is I don’t know, but the game theyre playing is a lot less about profits now. They care about control and power over people.
I theorize it has to do with desperation over what they see as an inevitable collapse of the United States and they are hedging their bets on holding onto the reigns of power for as long as possible until they can fuck off to their respective bunkers while the rest of humanity eats itself.
Then, when things settle they can peak their heads out of their hidie holes and start their new Utopian civilization or whatever.
Whatever’s going on, profits are not the focus right now. They are grasping at ways to control the masses…and failing pretty miserably I might add…though something tells me that scarcely matters to them.
inevitable collapse of the United States
Which they are intentionally trying to cause, rather that deal with their addiction to wealth and power.
Absolutely. It’s maddening that I’ve had to go from “maybe we should make society better somewhat” in my twenties to “if we’re gonna do capitalism, can we do it how it actually works instead of doing it stupid?” in my forties.
And the tragedy of the whole situation is that they can‘t win because if every worker is replaced by an algorithm or a robot then who‘s going to buy your products? Nobody has money because nobody has a job. And so the economy will shift to producing war machines that fight each other for territory to build more war machine factories until you can’t expand anymore for one reason or another. Then the entire system will collapse like the Roman Empire and we start from scratch.
producing war machines that fight each other for territory to build more war machine factories until you can’t expand anymore for one reason or another.
As seen in the retro-documentary Z!
Why would you need anyone to buy your products when you can just enjoy them yourself?
Nah, generative ai is pretty remarkably useful for software development. I’ve written dozens of product updates with tools like claudecode and cursorai, dismissing it as a novelty is reductive and straight up incorrect
As someone starting a small business, it has helped tremendously. I use a lot of image generation.
If that didn’t exist, I’d either has to use crappy looking clip art or pay a designer which I literally can’t afford.
Now my projects actually look good. It makes my first projects look like a highschooler did them last minute.
There are many other uses, but I rely on it daily. My business can exist without it, but the quality of my product is significantly better and the cost to create it is much lower.
Your product is other people’s work thrown in a blender.
Congrats.
Wait til you realize that’s just what art literally is…
You’re confusing ai art with actual art, like rendered from illustration and paintings
it’s as much “real” art as photography, taking a relatively finite number of decisions and finding something that looks “good”.
Really good photography is actually pretty hard and the best photographers are in high demand.
It involves a ton of settings for the camera, frequently post processing to balance out anything that wasn’t perfect during the shoot. Plus there is a ton of blocking, lighting, and if doing portraits and other planned shoots there is a lot of directing involved in getting the subjects to be in the right positions/showing the right emotions, etc. Even shooting nature requires a massive amount of planning and work beyond a few camera settings.
Hell, even stock photos tend to be a lot of work to set up!
If you think that someone taking a photo in focus with adequate lighting and posted it to instagram is the same as professional photography, then you have no idea what is involved.
Yeah he should be using real art like stock photos and shitty clip art
If their business can’t afford to pay someone qualified to do the work, the business shouldn’t exist.
I can stand by this for an established business. But we live in a capitalist society where you need money to make money. Until that changes, your ability to pay for work doesn’t have any bearing on the value of your new business venture.
I weep for your customers
As an experienced software dev I’m convinced my software quality has improved by using AI. More time for thinking and less time for execution means I can make more iterations of the design and don’t have to skip as many nice-to-haves or unit tests on account of limited time. It’s not like I don’t go through every code line multiple times anyway, I don’t just blindly accept code. As a bonus I can ask the AI to review the code and produce documentation. By the time I’m done there’s little left of what was originally generated.
As an experienced software dev I’m convinced my software quality has improved by using AI.
Then your software quality was extreme shit before. It’s still shit, but an improvement. So, yay “AI”, I guess?
That seems like just wishful thinking on your part, or maybe you haven’t learned how to use these tools properly.
Na, the tools suck. I’m not using a rubber hammer to get woodscrews into concrete and I’m not using “AI” for something that requires a brain. I’ve looked at “AI” suggestions for coding and it was >95% garbage. If “AI” makes someone a better coder it tells more about that someone than “AI”.
Then try writing the code yourself and ask ChatGPT’s o3-mini-high to critique your code (be sure to explain the context).
Or ask it to produce unit tests - even if they’re not perfect from the get go I promise you will save time by having a starting skeleton.
Another thing I often use it for is ad hoc transformations. For example I wanted to generate constants for all the SQLSTATE codes in the PostgreSQL documentation. I just pasted the table directly from the documentation and got symbolic constants with the appropriate values and with documentation comments.
As an experienced software dev, I know better than to waste my time writing boilerplate that can be vomited up by an LLM, since somebody else has already written it and I should just use that instead.
If a bot can develop your software better than you then you’re a shit software dev
That’s not what is happening. The bot writes code and then I tell it what to change until it’s close enough, then I make the final touches myself. It’s like having a junior programmer do the grunt work for you.
They’re all pretty fired up at the update velocity tbh 🤷
Yeah, nothing pleases us more than constant, buggy updates.
Unit tests and good architecture are still foundational requirements, so far no bug reports with any of these updates. In fact a huge chunk of these ai updates were addressing bugs. Not sure why you’re so mad at what you imagine is happening and making so many broad assumptions!
Don’t be an ass and realize that ai is a great tool for a lot of people. Why is that so hard to comprehend?
It’s not hard to comprehend. It’s that we literally have jackasses like Sam Altman arguing that if they can’t commit copyright violations at an industrial scale and pace that their business model falls apart. Yet, we’re still nailing regular people for piracy on an individual scale. As always individuals pay the price and are treated like criminals, but as long as you commit crime big enough and fast enough on an industrial scale, we shake our heads, go “wow” and treat you like a fucking hero.
If the benefits of this technology were evenly distributed the argument might have a leg to stand on, but it is never evenly distributed. It is always used as a way to pay professionals less for work that is “just okay.”
When a business buys the tools to use generative AI and they shitcan employees to afford it they have effectively used those employees labor against them to replace them with something lesser. Their labor was exploited to replace them. The people who actually deserve the bonus of generative AI are losing or being expected to be ten times more productive instead of being allowed to cool their heels because they worked hard enough to have this doohickey work for them. No, it’s always “line must go up, rich must get richer, fuck the laborers.”
I’ll stop being an ass about it when people stop burning employees out who already work hard or straight up fire them and replace them with this bullshit when their labor is what allowed the business to afford this bullshit to begin with. No manager or CEO can do all this labor on their own, but they get the fruits of all the labor their employees do as though they did do it all on their own, and it is fucked up.
I don’t have a problem with technology that makes our lives easier. I don’t have a problem with copyright violations (copyright as it exists is broken. It still needs to exist, just not in its current form).
What I have a problem with is businesses using this as an excuse to work their employees like slaves or replacing the employees that allowed them to afford these tools with these tools.
When everyone who worked hard to afford this stuff gets a paid vacation for helping to afford the tools and then comes back to an easier workload because the tools help that much, I’ll stop being a fucking ass about it.
Like I said elsewhere, the bottom line is business owners want a slave that doesn’t need things like sleep, food, emotional support, and never pushes back against being abused. I’m tired of people pretending like it’s not what businesses want. I’m tired of people pretending this does anything except make already overworked employees bust even more ass.
Your comment is on capitalism, not scaling ai or ai being used with effect.
What’s hard for you to comprehend about my comment?
You are insulting a person, because they said ai helps them.
😂
I don’t think any designer does work without heavily relying on ai. I bet that’s not the only profession.
It’s because customers don’t want it or care for it, it’s only the corporations themselves are obsessed with it
