this post was submitted on 27 Jan 2025
883 points (98.1% liked)
Technology
61206 readers
4373 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Here's someone doing 200 tokens/s (for context, OpenAI doesn't usually get above 100) on... A Raspberry Pi.
Yes, the "$75-$120 micro computer the size of a credit card" Raspberry Pi.
If all these AI models can be run directly on users devices, or on extremely low end hardware, who needs large quantities of top of the line GPUs?
Thank the fucking sky fairies actually, because even if AI continues to mostly suck it'd be nice if it didn't swallow up every potable lake in the process. When this shit is efficient that makes it only mildly annoying instead of a complete shitstorm of failure.
While this is great, the training is where the compute is spent. The news is also about R1 being able to be trained, still on an Nvidia cluster but for 6M USD instead of 500
True, but training is one-off. And as you say, a factor 100x less costs with this new model. Therefore NVidia just saw 99% of their expected future demand for AI chips evaporate
Even if they are lying and used more compute, it's obvious they managed to train it without access to the large amounts of the highest end chips due to export controls.
Conservatively, I think NVidia is definitely going to have to scale down by 50% and they will have to reduce prices by a lot, too, since VC and government billions will no longer be available to their customers.
It might also lead to 100x more power to train new models.
I doubt that will be the case, and I'll explain why.
As mentioned in this article,
This totally changes the way we think about AI training, which is why while OpenAI spent $100m on training GPT-4, running an expected 500,000 GPUs, DeepSeek used about 50,000, and likely spent that same roughly 10% of the cost.
So while operation, and even training, is now cheaper, it's also substantially less compute intensive to train models.
And not only is there less data than ever to train models on that won't cause them to get worse by regurgitating other worse quality AI-generated content, but even if additional datasets were scrapped entirely in favor of this new RL method, there's a point at which an LLM is simply good enough.
If you need to auto generate a corpo-speak email, you can already do that without many issues. Reformat notes or user input? Already possible. Classify tickets by type? Done. Write a silly poem? That's been possible since pre-ChatGPT. Summarize a webpage? The newest version of ChatGPT will probably do just as well as the last at that.
At a certain point, spending millions of dollars for a 1% performance improvement doesn't make sense when the existing model just already does what you need it to do.
I'm sure we'll see development, but I doubt we'll see a massive increase in training just because the cost to run and train the model has gone down.
Thank you. Sounds like good news.
I’m not sure. That’s a very static view of the context.
While china has an AI advantage due to wider adoption, less constraints and overall bigger market, the US has higher tech, and more funds.
OpenAI, Anthropic, MS and especially X will all be getting massive amounts of backing and will reverse engineer and adopt whatever advantages R1 had. Which while there are some it’s still not a full spectrum competitor.
I see the is as a small correction that the big players will take advantage of to buy stock, and then pump it with state funds, furthering the gap and ignoring the Chinese advances.
Regardless, Nvidia always wins. They sell the best shovels. In any scenario the world at large still doesn’t have their Nvidia cluster, think Africa, Oceania, South America, Europe, SEA who doesn’t necessarily align with Chinese interests, India. Plenty to go around.
Extra funds are only useful if they can provide a competitive advantage.
Otherwise those investments will not have a positive ROI.
The case until now was built on the premise that US tech was years ahead and that AI had a strong moat due to high computer requirements for AI.
We now know that that isn't true.
If high compute enables a significant improvement in AI, then that old case could become true again. But the prospects of such a reality happening and staying just got a big hit.
I think we are in for a dot-com type bubble burst, but it will take a few weeks to see if that's gonna happen or not.
Maybe, but there is incentive to not let that happen, and I wouldn’t be surprised if “they” have unpublished tech that will be rushed out.
The ROI doesn’t matter, it wasn’t there yet it’s the potential for it. The Chinese AIs are also not there yet. The proposition is to reduce FTEs, regardless of cost, as long as cost is less.
While I see OpenAi and mostly startups and VC reliant companies taking a hit, Nvidia itself as the shovel maker will remain strong.
if, on a modern gaming pc, you can get breakneck speeds of 5 tokens per second, then actually inference is quite energy intensive too. 5 per second of anything is very slow
That's becoming less true. The cost of inference has been rising with bigger models, and even more so with "reasoning models".
Regardless, at the scale of 100M users, big one-off costs start looking small.
But I’d imagine any Chinese operator will handle scale much better? Or?
Maybe? Depends on what costs dominate operations. I imagine Chinese electricity is cheap but building new data centres is likely much cheaper % wise than countries like the US.
Wth?! Like seriously.
I assume they are running the smallest version of the model?
Still, very impressive.
i can also run it on my old pentium from 3 decades ago. I'd have to swap 4MiB of weights in and out constantly, it will be very very slow, but it will work.
Sure you can run it on low end hardware, but how does the performance (response time for a given prompt) compare to the other models, either local or as a service?
That set of tokens/s is the performance, or response time if you'd like to call it that. GPT-o1 tends to get anywhere from 33-60, whereas in the example I showed previously, a Raspberry Pi can do 200 on a distilled model.
Now, granted, a distilled model will produce worse performance than the full one, as seen in a benchmark comparison done by DeepSeek here (I've outlined the most distilled version of the newest DeepSeek model, which is likely the kind that is being run on the Raspberry Pi, albeit likely with some changes made by the author of that post, as well as OpenAI's two most high-end models of a comparable distillation)
The gap in quality is relatively small for a model that is likely distilled far past what OpenAI's "mini" model is, when you consider that even regular laptop/PC hardware is orders of magnitudes more powerful than a Raspberry Pi, or that an external AI accelerator can be bought for as little as $60, the quality in practice could be very comparable with even slightly less distillation, especially with fine-tuning for a given use case (e.g. a local version of DeepSeek in a code development platform would be fine-tuned specifically just to produce code-related results)
If you get into the region of only cloud-hosted instances of DeepSeek that are running at-scale on GPUs like OpenAI's models are, the performance is only 1-2 percentage points off from OpenAI's model, at about 3-6% of the cost, which effectively means 3-6% of the total amount of GPU power being paid for compared to the amount of GPU power OpenAI is paying for.