6 ms·

ML is about hit another winter. Maybe intel is ahead of industry. Or we can keep asking high computers questions about programming.

by dgfitz 2y ago

PittleyDunkin 2y ago
> 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.
- klodolph 2y ago
  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.
- kimixa 2y ago
  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 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 https://www.bls.gov/news.release/ocwage.nr0.htm
  - BenjiWiebe 2y ago
    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.
    - semi-extrinsic 2y ago
      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.
      - kimixa 2y ago
        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.
  - choilive 2y ago
    I think the more accurate denominator would be the world population. People are seeing benefits to LLMs even outside of the office.
    - dgfitz 2y ago
      How do LLMs make money though?
    - hulitu 2y ago
      > 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)
      - 2y ago
- ben_w 2y ago
  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.
  - dgfitz 2y ago
    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.
- lm28469 2y ago
  > 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
  - PittleyDunkin 2y ago
    > 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.
seanmcdirmid 2y ago
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.
- vlovich123 2y ago
  What makes local AI interesting to you vs larger remote models like ChatGPT and Claude?
  - adriancr 2y ago
    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.
    - hentrep 2y ago
      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.
      - dgfitz 2y ago
        llama.ccp and time seems to be the general answer to this question.
      - adriancr 2y ago
        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...
HDThoreaun 2y ago
Selling cheap products that are worse than the competition is a valid strategy during downturns as businesses look to cut costs
throwaway48476 2y ago
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.

ML is about hit another winter. Maybe intel is ahead of industry. Or we can keep asking high computers questions about programming.

by dgfitz 2y ago

ML is about hit another winter. Maybe intel is ahead of industry. Or we can keep asking high computers questions about programming.

PittleyDunkin 2y ago

> 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.

klodolph 2y ago
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.
kimixa 2y ago
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 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 https://www.bls.gov/news.release/ocwage.nr0.htm
- BenjiWiebe 2y ago
  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.
  - semi-extrinsic 2y ago
    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.
    - kimixa 2y ago
      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.
- choilive 2y ago
  I think the more accurate denominator would be the world population. People are seeing benefits to LLMs even outside of the office.
  - dgfitz 2y ago
    How do LLMs make money though?
  - hulitu 2y ago
    > 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)
    - 2y ago
ben_w 2y ago
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.
- dgfitz 2y ago
  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.
lm28469 2y ago
> 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
- PittleyDunkin 2y ago
  > 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.

seanmcdirmid 2y ago

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.

vlovich123 2y ago
What makes local AI interesting to you vs larger remote models like ChatGPT and Claude?
- adriancr 2y ago
  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.
  - hentrep 2y ago
    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.
    - dgfitz 2y ago
      llama.ccp and time seems to be the general answer to this question.
    - adriancr 2y ago
      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...

HDThoreaun 2y ago

Selling cheap products that are worse than the competition is a valid strategy during downturns as businesses look to cut costs

throwaway48476 2y ago

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.