Buelldozer comments on Almost Half of US Data Centers That Were Supposed to Open This Year Slated to Be Canceled or Delayed

Almost Half of US Data Centers That Were Supposed to Open This Year Slated to Be Canceled or Delayed (futurism.com)

submitted 5 months ago by throws_lemy@reddthat.com to c/fuck_ai@lemmy.world

127 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[–] Buelldozer@lemmy.today -2 points 5 months ago* (last edited 5 months ago) (3 children)

NOBODY WANTS THIS SHIT!!!

That's a popular take, especially around here, but AI does have some pretty nice use cases; just not as many as the TechBros would have you believe.

Here's some examples I've personally seen in the last 14 days:

It's good at transcribing meetings, including picking out who is talking, backing into an agenda, and highlighting action items.
It's darn good at writing even moderately complex scripts in any of the common languages. (Powershell, Python, R, etc)
In the right hands (fingers?) it's getting increasingly good at finding and exploiting security flaws.
It's amazing at slicing and dicing data if the person using it knows what they're doing.

Does all of the "Agentic" Woo Woo shit work? No, it absolutely doesn't but it is clearly getting better as time goes on.

IMO this whole AI thing has some very strong parallels to the early '80s computer industry. Right now it often requires specialist knowledge for good results which makes it clunky to use, it is somewhat slow, there's very little interoperability, and it requires enormous amounts of power. Hell even this "over buying hardware" schtick fits right in, this happened with SRAM and then several times with DRAM as the industry matured.

However the industry is also making progress at almost insane speed; not only is the output getting demonstrably better but the negatives are being addressed. In the past 30 days I've seen prototype ASIC-esque hardware that works in a standard desktop PC and processes nearly 10,000 tokens a second with local processing.

The only reason you're not seeing that kind of kit in the market yet is because the models are still changing too much and no one wants to commit hundreds of millions to making cards that would be outdated before they could be shipped. We're probably only 18-24 months away though.

I've also seen 10x improvements in memory usage (TurboQuant) and literally dozens of little tweaks and tricks to reduce footprint and speed processing. Just like what was going on in the PC industry in the '80s and '90s.

So sure, Fuck AI (mostly) as it exists today but it won't be long before it's as ubiquitous as tablets and smartphones.

parent

hideshow 6 child comments

[–] MrKoyun@lemmy.world 18 points 5 months ago (1 child)

And we're fucking the world up to... transcribe meetings?

parent

hideshow 2 child comments

[–] OpenStars@piefed.social 9 points 5 months ago

No, it's to make the rich richer.

Many people do not think about what or why they are doing what they do, or what its end outcome will be.

parent

[–] Lost_My_Mind@lemmy.world 12 points 5 months ago (4 children)

I don't think you get why I don't want AI.

All the things you mentioned that AI is good at? Thats a bad thing to have. The more the technology becomes better, the worse all of our lives become.

AI will steal all jobs. ALL jobs. Even the prostitutes. Whatever your job is, AI within 10 years will do it better than you at a fraction of your cost. Basically for free. And you can't get another job, because ALL jobs are AI now. Build a robot, slap some AI in it, connect it to the main server, and it now has access to every AI units databases.

And then what about us? Well, the wealthy become the overlords, and we become the slaves.

parent

hideshow 8 child comments

[–] lemmy_outta_here@lemmy.world 3 points 5 months ago (1 child)

I actually agree with 99% of what you wrote, but you are a bit optimistic in one regard: they will want some sex slaves, but most of us will be food.

parent

hideshow 2 child comments

[–] Zink@programming.dev 3 points 5 months ago

Whoa whoa, has "eat the rich" been one of those situations where the hyphen/comma is in the wrong place?

It's really "Eat, the rich!"

parent

[–] bridgeburner@lemmy.world 2 points 5 months ago

How is AI gonna replace prostitutes? Maybe porn will shift to AI-generated stuff therefore reducing the number of porn actresses, but actual prostitutes? No way. How is AI gonna replace physical touch?

parent

[–] SummerReaper@lemmy.world 1 point 5 months ago

I think the only industry that's actually safe at this time is psychology. Therapy and mental health is bigger now than before. Plus it requires a real comprehensive understanding of the human experience that's simply impossible for AI to do effectively with positive results.

There probably be attempts though, I do think it'll be ruled as highly illegal.

parent

[–] Buelldozer@lemmy.today -1 points 5 months ago

You could be right, only time will tell.

parent

[–] aesthelete@lemmy.world 3 points 5 months ago (2 children)

So sure, Fuck AI (mostly) as it exists today but it won’t be long before it’s as ubiquitous as tablets and smartphones.

In order for it to be this ubiquitous it has to run locally or on commodity hardware IMO. The true lasting effects from this hype cycle are likely the capabilities that are being driven into smaller language models that don't have out of control resource requirements.

parent

hideshow 4 child comments

[–] Buelldozer@lemmy.today 2 points 5 months ago* (1 child)

In order for it to be this ubiquitous it has to run locally or on commodity hardware IMO.

I agree, which is why I shared that I recently saw a prototype ASIC-esque PCI card. The local hardware is coming, the models just need to settle down some before anyone will commit to building that hardware.

In the '90s and '00s you needed a zillion dollars of custom Silicon Graphics workstations and months of processing to do the FX for movies like "The Terminator". In 2020 you could replicate it in a few hours with commodity hardware.

The LLMs and AI will be the same, it just needs more than 5 years to get there.

parent

hideshow 2 child comments

[–] aesthelete@lemmy.world 2 points 5 months ago

Yeah if you can run them locally using a small board, that'll last.

parent

[–] boonhet@sopuli.xyz 1 point 5 months ago (1 child)

In order for it to be this ubiquitous it has to run locally or on commodity hardware IMO.

LLMs as they are, can already run on smartphones, which pretty are ubiquitous themselves.

So a flagship phone would have 12-16 gigs of RAM these days I believe. A low-end phone 4 gigs.

Here are the sizes of some different parameter count versions of Qwen 3.5, a popular Chinese open-weight LLM:

27B: 17 GB - not yet possible to run on current flagship phones, but once the RAM crisis ends, I could see this happening.

9B: 6.6 GB

4B: 3.4 GB

2B: 2.7 GB

0.8B: 1 GB.

For any recently manufactured device, there will be versions of multiple popular LLMs that will run on the RAM size they have available.

parent

hideshow 2 child comments

[–] aesthelete@lemmy.world 1 point 5 months ago* (1 child)

Most people do not have a smartphone with that amount of RAM. But ultimately, yeah, eventually it'll run on readily available hardware or it'll go into a dustbin.

There's already ollama and stuff. It'll stick around.

parent

hideshow 2 child comments

[–] boonhet@sopuli.xyz 1 point 5 months ago

I mean fairly low end phones are 4 GB now. They could likely afford running a model that fits in 1GB of RAM. Different models for different classes of phone even for the same manufacturer will likely be a thing.

parent