News for hackers. Stuff that matters.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by dgfitz on Tuesday December 03, 2024 @02:53PM (#42310534)
ML is about hit another winter. Maybe intel is ahead of industry.

Or we can keep asking high computers questions about programming.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by PittleyDunkin on Tuesday December 03, 2024 @02:57PM (#42310583)
> ML is about hit another winter.

I agree ML is about to hit (or has likely already hit) some serious constraints compared to breathless predictions of two years ago. I don't think there's anything equivalent to the AI winter on the horizon, though—LLMs even operated by people who have no clue how the underlying mechanism functions are still far more empowered than anything like the primitives of the 80s enabled.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by kimixa on Tuesday December 03, 2024 @03:10PM (#42310729)
I think there'll be a "financial" winter - or another way a bubble burst - the investment right now is simply unsustainable, how are these products going to be monetized?

Nvidia had a revenue of $27billion in 2023 - that's about $160 per person per year [0] for every working age person in the USA. And it's predicted to more than double in 2024. If you reduce that to office workers (you know, the people who might actually get some benefit, as no AI is going to milk a cow or serve you starbucks) that's more like $1450/year. Or again more than double that for 2024.

How much value add is the current set of AI products going to give us? It's still mostly promise too.

Sure, like most bubbles there'll probably still be some winners, but there's no way the current market as a whole is sustainable.

The only way the "maximal AI" dream income is actually going to happen is if they functionally replace a significant proportion of the working population completely. And that probably would have large enough impacts to society that things like "Dollars In A Bank" or similar may not be so important.

[0] Using the stat of "169.8 million people worked at some point in 2022" https://www.bls.gov/news.release/pdf/work.pdf

[1] 18.5 million office workers according to https://www.bls.gov/news.release/ocwage.nr0.htm

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by ben_w on Tuesday December 03, 2024 @03:27PM (#42310936)
While I'd agree monetisation seems to be a challenge in the long term (analogy: spreadsheets are used everywhere, but are so easy to make they're not themselves a revenue stream, only as part of a bigger package)…

> Nvidia had a revenue of $27billion in 2023 - that's about $160 per person per year [0] for every working age person in the USA

As a non-American, I'd like to point out we also earn money.

> as no AI is going to milk a cow or serve you starbucks

Cows have been getting the robots for a while now, here's a recent article: https://modernfarmer.com/2023/05/for-years-farmers-milked-co...

Robots serve coffee as well as the office parts of the coffee business: https://www.techopedia.com/ai-coffee-makers-robot-baristas-a...

Some of the malls around here have food courts where robots bring out the meals. I assume they're no more sophisticated than robot vacuum cleaners, but they get the job done.

Transformer models seem to be generally pretty good at high-level robot control, though IIRC a different architecture is needed down at the level of actuators and stepper motors.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by kimixa on Tuesday December 03, 2024 @04:38PM (#42311820)
Sure, robotics help many jobs, and some level of the current deep learning boom seems to have crossover in improving that - but how many of them are running LLMs that affect Nvidia's bottom line right now? There's some interesting research in that area, but it's certainly not the primary driving force. And then is the control system the limiting factor for many systems - it's probably relatively easy to get a machine today that makes a Starbucks coffee "as good as" a decently trained human. But the market doesn't seem to want that.

And I know restricting it to the US is a simplification, but so is restricting it to Nvidia, it's just to give a ballpark back-of-the-envelope "does this even make sense?" level calculation. And that's what I'm failing to see.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by amluto on Tuesday December 03, 2024 @05:35PM (#42312366)
Machines that will make espresso, automatically, that I personally like better than what Starbucks serves are widely available. No AI needed, and they aren't even "robotic". They can use ordinary coffee beans, and you can get them for home use or for commercial use. You can also go to a mall and get a robot to make you coffee.

Nonetheless, Starbucks does not use these machines, and I don't see any reason that AI, on its current trajectory, will change that calculation any time soon.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by lm28469 on Tuesday December 03, 2024 @05:37PM (#42312388)
I love how the fact that we might not want AI/robots everywhere in our lives isn't even discussed.

They could serve us a plate of shit and we'd debate if pepper or salt is better to complement it

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by ben_w on Tuesday December 03, 2024 @05:47PM (#42312495)
It's pretty often discussed, it's just hard to put everything into a single comment (or thread).

I mean, Yudkowsky has basically spent the last decade screaming into the void about how AI will with high probability literally kill everyone, and even people like me who think that danger is much less likely still look at the industrial revolution and how slow we were to react to the harms of climate change and think "speed-running another one of these may be unwise, we should probably be careful".

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by choilive on Tuesday December 03, 2024 @03:22PM (#42310875)
I think the more accurate denominator would be the world population. People are seeing benefits to LLMs even outside of the office.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by dgfitz on Tuesday December 03, 2024 @04:08PM (#42311454)
How do LLMs make money though?
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by hulitu on Tuesday December 03, 2024 @04:08PM (#42311457)
> I think the more accurate denominator would be the world population. People are seeing benefits to LLMs even outside of the office.

For example ?

(besides deep fakes)

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by ben_w on Wednesday December 04, 2024 @05:16AM (#42316106)
I use one in the kitchen, because it's easier than the ads and prose on most recipe websites, and it can adapt to whatever ingredients I actually have rather than being a fixed list.

Used them in the garden while weeding and in the garden store while planning what to plant, in both cases to identify plants by image and tell me about them — though I'd say image capable AI are no longer mere "large language models".

Used ChatGPT while shopping to help me locate products in the store I was in, when I couldn't find just by wandering the isles, by uploading a photo of the aisle I happened to be in at the point I gave up.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by BenjiWiebe on Tuesday December 03, 2024 @03:20PM (#42310853)
Well, "AI" is milking cows. Not LLM's though. Our milking robot uses image recognition to find the cow's teats to put the milking cup on.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by semi-extrinsic on Tuesday December 03, 2024 @03:26PM (#42310924)
Yeah, but automated milking robots like that have been in the market for more than a decade now IIRC?

Seems like a lot of CV solutions have seen fairly steady but small incremental advances over the past 10-15 years, quite unrelated to the current AI hype.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by kimixa on Tuesday December 03, 2024 @04:42PM (#42311865)
Improving capabilities of AI isn't at odds with expecting an "AI Winter" - just the current drive is more hype than sustainable, provable progress.

We've been through multiple AI Winters, as a new technique is developed, it does increase the capabilities. Just not as much as the hype suggested.

To say there won't be a bust implies this boom will last forever, into whatever singularity that implies.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by ben_w on Tuesday December 03, 2024 @03:15PM (#42310782)
What we had in the 80s was barely able to perform spell-check, free downloadable LLMs today are mind-blowing even in comparison to GPT-2.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by dgfitz on Tuesday December 03, 2024 @04:09PM (#42311473)
I think the only good thing that came out of the 80s was the 90s. I’d leave that decade alone so we can forget about it.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by klodolph on Tuesday December 03, 2024 @03:03PM (#42310645)
Yeah… I want to think of it like mining, where you’ve found an ore vein. You have to switch from prospecting to mining. There’s a lot of work to be done by integrating our LLMs and other tools with other systems, and I think the cost/benefit of making models bigger, Bigger, BIGGER is reaching a plateau.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by lm28469 on Tuesday December 03, 2024 @05:32PM (#42312337)
> even operated by people who have no clue how the underlying mechanism functions are still far more empowered than anything like the primitives of the 80s enabled.

I'm still not convinced about that. All the """studies""" show 30-60% boost in productivity but clearly this doesn't translate to anything meaningful in real life because no industry laid off 30-60% of their workforce and no industry progressed anywhere close to 30% since chat gpt was released.

It's been released a whole 24 months ago, remember the talks about freeing us from work and curing cancer... Even investments funds which are the biggest suckers for anything profitable are more and more doubtful

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by PittleyDunkin on Wednesday December 04, 2024 @12:43PM (#42319890)
> All the """studies""" show 30-60% boost in productivity

The services that provide serious productivity boosts aren't being heavily used or marketed yet. They:

1. Attempt to do narrow tasks with high proficiency. 2. Replace specific job titles. 3. Are high-value enough to be slower to engage in layoffs.

What I'm suggesting is not a fungible solution, but it is one that will be highly profitable and productive.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by seanmcdirmid on Tuesday December 03, 2024 @03:03PM (#42310643)
Haven't people been saying that for the last decade? I mean, eventually they will be right, maybe "about" means next year, or maybe a decade later? They just have to stop making huge improvements for a few years and the investment will dry up.

I really wasn't interested in computer hardware anymore (they are fast enough!) until I discovered the world of running LLMs and other AI locally. Now I actually care about computer hardware again. It is weird, I wouldn't have even opened this HN thread a year ago.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by vlovich123 on Tuesday December 03, 2024 @03:05PM (#42310660)
What makes local AI interesting to you vs larger remote models like ChatGPT and Claude?
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by adriancr on Tuesday December 03, 2024 @03:09PM (#42310713)
Not OP but for me a big thing is privacy, I can feed it personal documents and expect those to not leak.

It has zero cost, hardware is already there. I'm not captive to some remote company.

I can fiddle and integrate with other home sensors / automation as I want.

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by hentrep on Tuesday December 03, 2024 @03:19PM (#42310823)
Curious as I’m of the same mind - what’s your local AI setup? I’m looking to implement a local system that would ideally accommodate voice chat. I know the answer depends on my use case - mostly searching and analysis of personal documents - but would love to hear how you’ve implemented.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by dgfitz on Tuesday December 03, 2024 @03:57PM (#42311317)
llama.ccp and time seems to be the general answer to this question.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by adriancr on Wednesday December 04, 2024 @04:29AM (#42315868)
If you are just starting up, you can try out 'open-webui' as inspiration. After that you can just use llama.cpp to build out your own things.

Hardware side, I just have a beefy server that acts as a router (mellanox card to provider fiber optic and local fiber network), firewall, wifi access point, zigbee coordinator, host to various services, camera video feed ingestion and processing, and so on...

Re:Intel announces Arc B-series "Battlemage" discrete graphi... by red-iron-pine on Wednesday December 04, 2024 @10:35AM (#42318452)
define brefy server
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by seanmcdirmid on Tuesday December 03, 2024 @03:50PM (#42311238)
Control and freedom. You can use unharmonious models and hacks to existing models, also latency, you can actually use AI for a lot more applications when it is running locally.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by epicureanideal on Tuesday December 03, 2024 @03:09PM (#42310716)
Lack of ideological capture of the public models.
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by HDThoreaun on Tuesday December 03, 2024 @03:17PM (#42310800)
Selling cheap products that are worse than the competition is a valid strategy during downturns as businesses look to cut costs
Re:Intel announces Arc B-series "Battlemage" discrete graphi... by throwaway48476 on Tuesday December 03, 2024 @04:25PM (#42311665)
The survivors of the AI winter are not the dinosaurs but the small mammals that can profit by dramatically reducing the cost of AI inference in a minimum Capex environment.