Deepseek is good locally.
this post was submitted on 28 Jan 2025
277 points (89.2% liked)
memes
16604 readers
2265 users here now
Community rules
1. Be civil
No trolling, bigotry or other insulting / annoying behaviour
2. No politics
This is non-politics community. For political memes please go to !politicalmemes@lemmy.world
3. No recent reposts
Check for reposts when posting a meme, you can only repost after 1 month
4. No bots
No bots without the express approval of the mods or the admins
5. No Spam/Ads
No advertisements or spam. This is an instance rule and the only way to live.
A collection of some classic Lemmy memes for your enjoyment
Sister communities
- !tenforward@lemmy.world : Star Trek memes, chat and shitposts
- !lemmyshitpost@lemmy.world : Lemmy Shitposts, anything and everything goes.
- !linuxmemes@lemmy.world : Linux themed memes
- !comicstrips@lemmy.world : for those who love comic stories.
founded 2 years ago
MODERATORS
As someone who is rather new to the topic: I have a GPU with 16 GB VRAM and only recently installed Ollama. Which size should I use for Deepseek R1?🤔
You can try from lowest to bigger. You probably can run biggest too but it will be slow.
I also have 16gb vram and the 32b version runs ok. Anything larger would take too long I think
Every sane person working in tech saw that coming, lol. Not saying that AI is bullshit, but it doesn't do anything that marketing want you to believe.
100%, AI moves in cycles related to scientific research, same as most other tech innovation funding hype boom/bust cycles.
Incremental scientific research delivers new possible features -> funding hype to be first to market -> engineering builds products based on new tech -> people learn what it can and cannot do through use -> hype dies and funding dries up (you are here for LLMs) -> less funding stops engineering new products and research continues incremental changes until repeat.
It's been that way for AI since the 80s with expert systems, then deep nets, and now LLMs. I suspect a lot of layoffs over the past year have been related to over committing to commercial office space and AI solutions not delivering projected ROIs.
Incremental scientific research delivers new possible features -> funding hype to be first to market -> engineering builds products based on new tech -> people learn what it can and cannot do through use -> hype dies and funding dries up (you are here for LLMs) -> less funding stops engineering new products and research continues incremental changes until repeat.
TLDR: AI works on goverment funding.
10-15 years ago it was cloud everything. Now it's AI. Gotta have a new hype to draw in the VC money. It's a lot of the same players who got rich on the last hype cycle. Cloud was absolutely not bullshit and neither is AI. But like 90 percent of the claims are marketing fluff.
What am I looking at? What's the reason behind this collapse?
My horrible take only based on a few snippets:
Chinese open source AI releases, wipes the floor with US AI. Exposes AI as massively overvalued. Not surprising considering how big tech is just taking everything. Nvidia loses like half a trillion in value.
I’m happy to see a home run against the fascists that hurt them more than violence, at least to start. And I’m not a fan of AI (as hyped), but if it’s going to happen I’d rather it be with an open source performer as a base than the venture capitalists that abandoned us.
And really the US is no more trustworthy than China for those of us under the shadow of the big dick energy stuporpowers.
That open source performer is very much aligned with Winnie the poohs censorship, which should be worrying.
The model has no censorship. Some websites running the model are hosted in China and thus do. Use one of the thousands of instances that run out of the US or run it yourself. It's not like chatgpt or other openai trash, it's system requirements are pretty low.
One fascism vs. the other fascism, and everyone loses.
Home run against the fascists, in favor of the other fascists
At least red fascists get healthcare. American fascists are really going to need to step up their game on the bread and circus front.
I would honestly be more pro China if they weren’t so unbearably heavy on the censorship front
load more comments
(4 replies)
load more comments
(9 replies)
Not overvalued exactly. I believe the main point is they claimed to have used much much less power to train it, and that it demonstrates equivalent capability on much much less power.
There’s been a huge stock bubble on speculation of near infinite demand for datacenters, and NVidia processors. If DeepSeek’s claims are true, they just popped that bubble. We don’t suddenly need so many new datacenters. Demand for NVidia processors is not infinite, and their stock price is excessive compared to reality
Chinese ChatGPT outperforms American models, on cheaper and less powerful chips as well. Leading to speculations of overvalued tech stocks in the US on an already very high valued market
load more comments
(2 replies)
It’s not a collapse (yet). This is the first of many market corrections as AI drops back into the realm of reality.
Loss in trust in the US economy under Trump along with the appearance of a new "Hail Mary" by investors who can look outside of the US' national borders unlike Trump and his followers.
You know, I used to worry about China taking over the world, but honestly they look like a much better choice for a super power than the US is now.
I for one welcome our new communist overlords.
It'll require us to create an actual AI first, but yeah when that day comes shit's gonna get real wild real fast.
And no, the glorified spellchecker-on-steroids we're all calling AI but isn't actual intelligence isn't gonna cut it. Fun toy though.
glorified spellchecker-on-steroids
What?
glorified spellchecker-on-steroids
It’s more of a predictive keyboard on steroids.
Outlook's integration of said concept just made me copy and paste a sentence from notepad because it repeatedly tried to correct a phrase and wouldn't fucking stop doing it.
Because who the fuck would ever type the letters "my" and then follow it with a space?
It's artificial as in fake, not man-made and it's been that way since the 70's. It's only the pop-sci definition that's inaccurate. AI has always been about faking decision making whether it is an adversarial AI in a video game or a chat bot. You can hate the overhyped wallstreet buzz around LLMs, but this semantic argument crap has just been annoying.
It’s artificial as in fake, not man-made and it’s been that way since the 70’s
'Artificial' isn't the part of it that's in question, lol. And computing has been around since way before the 70s.
AI has always been about faking decision making
That isn't intelligence.
adversarial AI in a video game or a chat bot
are examples of A but not I.
this semantic argument crap has just been annoying.
Agreed. Words have meanings, and while we fuck it up a lot by not understanding the meaning or by some marketing department deliberately using the wrong words to sell something to people who don't know any better, none of that means the words don't have meanings.
Unless we're going with the whole "we've used it incorrectly long enough that that's what it means now cuz language evooooolves!" shit, and honestly, maybe we have crossed that line, but that shit's above my paygrade.
load more comments
(2 replies)
After this one?
The techbro authoritarian society is coming.
What application is this?
CHAINA
I read that in Trump’s voice lol
Is now a good time to invest in stocks?
Do you have money to throw away?
The best time is yesterday. The second best time is today.
Now is just as speculative a time to buy as yesterday. Stocks are unstable. Maybe someone will prove DeepSeek lied and stocks will explode again. Maybe they’ll be proven true and stocks will drop much more.
Buy the dip
Funnily enough, this actually worked the first time GME crashed.
Just not the second.
Buy the dip, not the others.
load more comments
(4 replies)
Bur I don't write Mandarin.
Don't worry they write in English for you
view more: next ›