5 ms·

"I'm generally an AI skeptic, but it seems awfully early to make this call." What call. Maybe some readers miss the (perhaps subtle) difference between "Gener

"I'm generally an AI skeptic, but it seems awfully early to make this call." What call. Maybe some readers miss the (perhaps subtle) difference between "Generative AI is not ..." and "Generative Ai is not going to ..." Then first can be based on fact, e.g., what has happened so far. The second is based on pure speculation. No one knows what will happen in the future. HN is continually being flooded with speculation, marketing, hype. In contrast, this article, i.e., the paper it discusses, is is based on what has happened so far. There is no "call" being made. Only an examination of what has heppened so far. Facts not opinions.

mac-attack 2y ago
> In contrast, this article, i.e., the paper it discusses, is is based on what has happened so far. What happened in 2023 and 2024 actually Nitpicky but it's worth noting that last year's AI capabilities are not the April 2025 AI capabilities and definitely won't be the December 2025 capabilities. It's using deprecated/replaced technology to make a statement, that is not forward projecting. I'm struggling to see the purpose. It's like announcing that the sun is still shining at 7pm, no?
- rafaelmn 2y ago
  I feel like model improvement is severely overstated by the benchmarks and the last release cycle basically made no difference to my use cases. If you gave me Claude 3.5 and 3.7 I couldn't really tell the difference. OpenAI models feel like they are regressing, and LLAMA 4 regressed even on benchmarks. And the hype was insane in 2023 already - it's useful to compare actual outcomes vs historic hype to gauge how credible the hype sellers are.
  - kypro 2y ago
    That's interesting. I think there's been some pretty significant improvements in the rate of hallucinations and accuracy of the models, especially when it comes to rule following. Perhaps the biggest improvement though is in the size of context windows which are huge compared to this time last year. Maybe progress over the last 2-3 months is hard to see, but progress over the last 6 is very clear.
- freejazz 2y ago
  Well, it's a given that the sun is shining when it is out. Not so much with the AI.
bredren 2y ago
> According to the study, "users report average time savings of just 2.8 percent of work hours" from using AI tools. That's a bit more than one hour per 40 hour work week. Could be data is lagging as sibling comment said but this seems wildly difficult to report on a number like this. It also doesn't take into account the benefits to colleagues of active users of LLMs (second order savings). My use of LLMs often means I'm saving other people time because I can work through issues without communications loops and task switching. I can ask about much more important, novel items of discussion. This is an important omission that lowers the paper's overall value and sets it up for headlines like this.
- jaredklewis 2y ago
  The headline is about wages and jobs. It’s very possible that AI could result in time savings of 50% of work hours in a week and still have no impact on wages or jobs. This is because the economy is not a static thing. If one variable changes (productivity), it’s not a given that GDP will remain constant and jobs/wages will consequently be reduced. More likely is that all of the variables are always in flux, reacting and responding to changes in the market.
  - bredren 2y ago
    Very well, I acknowledge this point. However, the parent comment is about an examination of what has happened so far and facts that feed into the paper and its conclusions. I was focused on what I see as important gaps in measuring impact of AI, and its actual (if difficult to measure) impact right now.
    - jaredklewis 2y ago
      The paper analyzes facts re: wages and jobs, which I think are (comparatively) easy to measure as compared with productivity, and are also an area where people have concerns about the impact of AI. Mostly people aren't worried about productivity itself, which would be weird. "Oh no, AI is making us way more productive, and now we're getting too much stuff done and the economy is growing too much." The major concern is that the productivity is going to impact jobs and wages, and at least so far (according to this particular paper) that seems to not be happening.
hn_throwaway_99 2y ago
> "Generative AI is not ..." But here's the thing - there is already plenty of documented proof of individuals losing their job to ChatGPT. This is an article from 2 years ago: https://www.washingtonpost.com/technology/2023/06/02/ai-taking-jobs/ https://www.washingtonpost.com/technology/2023/06/02/ai-taki... Early on in a paradigm shift, when you have small moves, or people are still trying to figure out the tech, it's likely that individual moves are hard to distinguish from noise. So I'd argue that a broad-based, "just look at the averages" approach is simply the wrong approach to use at this point in the tech lifecycle. FWIW, I'd have to search for it, but there were economic analyses done that said it took decades for the PC to have a positive impact on productivity. IMO, this is just another article about "economists using tools they don't really understand". For decades they told us globalization would be good for all countries, they just kinda forgot about the massive political instability it could cause. > In contrast, this article, i.e., the paper it discusses, is based on what has happened so far. Not true. The article specifically calls into question whether the massive spending on AI is worth it. AI is obviously an investment, so determine whether it's "worth it", you need to consider future outcomes.
- rightbyte 2y ago
  > economic analyses done that said it took decades for the PC to have a positive impact on productivity I honestly think computers have a net negative productivity impact in many organizations. Maybe even "most".
  - 7qW24A 2y ago
    It’s an interesting rabbit hole to go down. If you use the BLS’s definition of productivity, then computers seem to be a net drag on productivity: https://usafacts.org/articles/what-is-labor-productivity-and-how-has-it-changed-in-the-us-over-time/ https://usafacts.org/articles/what-is-labor-productivity-and... Even more surprising for me is that productivity growth declined during the ZIRP era. How did we take all that free money and product less?
    - Arctic_fly 2y ago
1vuio0pswjnm7 2y ago
Whether "AI" investements are "worth it" is not a question considered by the paper. The question it addresses is whether "AI chatbots have had a significant impact on earnings or recorded hours in an occupation." The paper makes no "call". It estimates that so far, there has been no significant impact. This is not a forecast or a prediction. It is a review of past results.
edanm 2y ago
You've got a point. But I think it's pretty clear what the parent poster means - that it's too early to imply that GenAI won't impact jobs. Obviously you understood that that was the point parent wanted to make - because you argue against it.

mac-attack 2y ago

> In contrast, this article, i.e., the paper it discusses, is is based on what has happened so far. What happened in 2023 and 2024 actually Nitpicky but it's worth noting that last year's AI capabilities are not the April 2025 AI capabilities and definitely won't be the December 2025 capabilities. It's using deprecated/replaced technology to make a statement, that is not forward projecting. I'm struggling to see the purpose. It's like announcing that the sun is still shining at 7pm, no?

rafaelmn 2y ago
I feel like model improvement is severely overstated by the benchmarks and the last release cycle basically made no difference to my use cases. If you gave me Claude 3.5 and 3.7 I couldn't really tell the difference. OpenAI models feel like they are regressing, and LLAMA 4 regressed even on benchmarks. And the hype was insane in 2023 already - it's useful to compare actual outcomes vs historic hype to gauge how credible the hype sellers are.
- kypro 2y ago
  That's interesting. I think there's been some pretty significant improvements in the rate of hallucinations and accuracy of the models, especially when it comes to rule following. Perhaps the biggest improvement though is in the size of context windows which are huge compared to this time last year. Maybe progress over the last 2-3 months is hard to see, but progress over the last 6 is very clear.
freejazz 2y ago
Well, it's a given that the sun is shining when it is out. Not so much with the AI.

bredren 2y ago

> According to the study, "users report average time savings of just 2.8 percent of work hours" from using AI tools. That's a bit more than one hour per 40 hour work week. Could be data is lagging as sibling comment said but this seems wildly difficult to report on a number like this. It also doesn't take into account the benefits to colleagues of active users of LLMs (second order savings). My use of LLMs often means I'm saving other people time because I can work through issues without communications loops and task switching. I can ask about much more important, novel items of discussion. This is an important omission that lowers the paper's overall value and sets it up for headlines like this.

jaredklewis 2y ago
The headline is about wages and jobs. It’s very possible that AI could result in time savings of 50% of work hours in a week and still have no impact on wages or jobs. This is because the economy is not a static thing. If one variable changes (productivity), it’s not a given that GDP will remain constant and jobs/wages will consequently be reduced. More likely is that all of the variables are always in flux, reacting and responding to changes in the market.
- bredren 2y ago
  Very well, I acknowledge this point. However, the parent comment is about an examination of what has happened so far and facts that feed into the paper and its conclusions. I was focused on what I see as important gaps in measuring impact of AI, and its actual (if difficult to measure) impact right now.
  - jaredklewis 2y ago
    The paper analyzes facts re: wages and jobs, which I think are (comparatively) easy to measure as compared with productivity, and are also an area where people have concerns about the impact of AI. Mostly people aren't worried about productivity itself, which would be weird. "Oh no, AI is making us way more productive, and now we're getting too much stuff done and the economy is growing too much." The major concern is that the productivity is going to impact jobs and wages, and at least so far (according to this particular paper) that seems to not be happening.

hn_throwaway_99 2y ago

> "Generative AI is not ..." But here's the thing - there is already plenty of documented proof of individuals losing their job to ChatGPT. This is an article from 2 years ago: https://www.washingtonpost.com/technology/2023/06/02/ai-taking-jobs/ https://www.washingtonpost.com/technology/2023/06/02/ai-taki... Early on in a paradigm shift, when you have small moves, or people are still trying to figure out the tech, it's likely that individual moves are hard to distinguish from noise. So I'd argue that a broad-based, "just look at the averages" approach is simply the wrong approach to use at this point in the tech lifecycle. FWIW, I'd have to search for it, but there were economic analyses done that said it took decades for the PC to have a positive impact on productivity. IMO, this is just another article about "economists using tools they don't really understand". For decades they told us globalization would be good for all countries, they just kinda forgot about the massive political instability it could cause. > In contrast, this article, i.e., the paper it discusses, is based on what has happened so far. Not true. The article specifically calls into question whether the massive spending on AI is worth it. AI is obviously an investment, so determine whether it's "worth it", you need to consider future outcomes.

rightbyte 2y ago
> economic analyses done that said it took decades for the PC to have a positive impact on productivity I honestly think computers have a net negative productivity impact in many organizations. Maybe even "most".
- 7qW24A 2y ago
  It’s an interesting rabbit hole to go down. If you use the BLS’s definition of productivity, then computers seem to be a net drag on productivity: https://usafacts.org/articles/what-is-labor-productivity-and-how-has-it-changed-in-the-us-over-time/ https://usafacts.org/articles/what-is-labor-productivity-and... Even more surprising for me is that productivity growth declined during the ZIRP era. How did we take all that free money and product less?
  - Arctic_fly 2y ago

1vuio0pswjnm7 2y ago

Whether "AI" investements are "worth it" is not a question considered by the paper. The question it addresses is whether "AI chatbots have had a significant impact on earnings or recorded hours in an occupation." The paper makes no "call". It estimates that so far, there has been no significant impact. This is not a forecast or a prediction. It is a review of past results.

edanm 2y ago

You've got a point. But I think it's pretty clear what the parent poster means - that it's too early to imply that GenAI won't impact jobs. Obviously you understood that that was the point parent wanted to make - because you argue against it.