3 ms·

If you don’t go fast you deprive your consumers of your product. It’s not clear why that tradeoff is good? There was a recent petition signed by all major AI l

If you don’t go fast you deprive your consumers of your product. It’s not clear why that tradeoff is good? There was a recent petition signed by all major AI labs to slow down AI development. Would this author or you guys agree it’s a good thing?

dgellow 3mo ago
I would, yes, given how unsustainable the whole AI industry looks like. It would be way, way better if they could figure out their things out before pushing so hard for the whole software world to adopt their experimental tech
- simianwords 3mo ago
  if they took their time then billions of people wouldn't have access to it for years and decades. its now super clear that AI is a step improvement in coding, mathematics and other domains. the push for AI has worked out in hindsight. there will always be people who will parrot the METR study and claim productivity didn't increase but its best to ignore them.
  - dgellow 3mo ago
    I don’t see the problem with the technology not being available widely for way longer as it improves. If you think the current situation is a good one you’re not paying attention to how ridiculous the spending has been on the AI bet. What exists now is not sustainable at all, and is already the cause of a massive worldwide inflation. And so far no proof of improving companies ROI. The AI push is going to damage our societies for a long time
    - simianwords 3mo ago
      Why do you think the spending is ridiculous? Do you not agree that we've had a step improvement in coding, mathematics, search/retrieval?
      - dgellow 3mo ago
        It's irrelevant... The spending commitment OpenAI has for 2030 is larger than the projected AI infra market in 2030. And we still haven't seen a proof that agents contribute positively to consumers ROI. It's a capital misallocation problem, a number problem, that has nothing to do with subjective feelings regarding LLMs
        simianwords 3mo ago
        In that case, if the numbers do work out and labs make shit load of money, would you agree you were mistaken?
        dgellow 3mo ago
        no... please actually look at the information we know about AI vendors and how much debt they have. OpenAI is projected to have more than $20B in losses for 2026. They wouldn't _just_ need to make a shit ton of money, they would need to first become profitable, then do a shit ton of money while staying profitable, for multiple years, just to match their spending commitment. That's not even taking in account their ROC and company valuation
        simianwords 3mo ago
        Sure but if they do make money and profit, your concern would have been invalid in hindsight wouldn’t it? Your concern seems to be profitability of OpenAI but that’s easily falsifiable. I’m not saying it’s 100% but I’m saying if they do turn out profitable, your concern wouldn’t have been valid.
        dgellow 3mo ago
        depends on the exact chain of events, but if the economics are indeed solved, and AI labs are found to have a sustainable business model, that concern would be addressed. I would still see the technology as anti-human and would still see the concept of an agentic economy as deeply unserious, wasteful, and risky. Not exactly sure what your point is though. FWIW my concerns aren't only for OpenAI, it's just the poster-child of the AI bubble. SpaceX AI strategy and investment is a complete joke (the S-1 they filled is an insult to a reader's intelligence). Anthropic seems to have been more cautious but has the same fundamental economic problems. Nobody in that whole industry has a moat, the top AI vendors are way too exposed to Chinese/open-weights labs. The datacenters debt investment vehicles (SPVs) look extremely shady and made to obfuscate the underlying assets. The HBM manufacturers seem to be going through one of the most violent boom-burst cycle ever. Oracle is... doing Oracle things, I would be shocked if that company is still alive in its current form in 5y. And so on
rglover 3mo ago
I'm the author. If you do go fast and don't deprive your customers of your product but the product breaks and doesn't deliver its intended value, is that a better tradeoff? Re: AI development, I'd say it's foolish to slow down research, but worthwhile for rollouts to the market. The time that should have been used for preparing people/businesses was wasted on fearmongering campaigns. Now we have a fire hose of slop that's out of control. Why? Rushing.
- simianwords 3mo ago
  > If you do go fast and don't deprive your customers of your product but the product breaks and doesn't deliver its intended value, is that a better tradeoff? Its not a good tradeoff if it breaks it enough that people move off your product, I obviously concede that. Ultimately a firm should be incentivised to improve their long term profits after accounting for externalities to the world. Fast or slow is an implementation detail. >Re: AI development, I'd say it's foolish to slow down research, but worthwhile for rollouts to the market. The time that should have been used for preparing people/businesses was wasted on fearmongering campaigns. Now we have a fire hose of slop that's out of control. Its not clear that your subjective view of slop is reason enough for the world to slow down on AI rollout and I'm glad I don't live in a world that doesn't necessitate slowing down due to whims and fancies of a few people's subjective aesthetic takes. Ironically for me, all the companies _do_ want to slow down their pace for reasons related to actual safety. I think you would appreciate the cause and the foresight (if any). https://www.pacingthefrontier.com/ https://www.pacingthefrontier.com/
  - rglover 3mo ago
    > Its not clear that your subjective view of slop is reason enough for the world to slow down on AI rollout and I'm glad I don't live in a world that doesn't necessitate slowing down due to whims and fancies of a few people's subjective aesthetic takes. This is a misunderstanding of what I'm talking about. It's not about "slop" in and of itself, it's about what reality that slop inevitably creates. There's a lot of focus on the output ("ermagherd you wrote this with AI"), but very little on the ripple effects/second-order effects (people spinning up cults, ending marriages, and gambling away their money [1]). That's not just cute "look at the freaks" stuff, that's a serious, civilizational-level problem 1-2 decades out. You're right that my subjective opinion means dick-all beyond being that of an experienced, educated user of this stuff. But I'd at least hope that people take heed of what I'm saying and start to push back when it's objectively clear that decisions are closer on the spectrum to disaster than they are to benign mediocrity. Thank you for sharing that link, this is exactly what I'd hope to see happening behind the scenes. [1] https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chatbot-users-lives-wrecked-by-delusion https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chat...

dgellow 3mo ago

I would, yes, given how unsustainable the whole AI industry looks like. It would be way, way better if they could figure out their things out before pushing so hard for the whole software world to adopt their experimental tech

simianwords 3mo ago
if they took their time then billions of people wouldn't have access to it for years and decades. its now super clear that AI is a step improvement in coding, mathematics and other domains. the push for AI has worked out in hindsight. there will always be people who will parrot the METR study and claim productivity didn't increase but its best to ignore them.
- dgellow 3mo ago
  I don’t see the problem with the technology not being available widely for way longer as it improves. If you think the current situation is a good one you’re not paying attention to how ridiculous the spending has been on the AI bet. What exists now is not sustainable at all, and is already the cause of a massive worldwide inflation. And so far no proof of improving companies ROI. The AI push is going to damage our societies for a long time
  - simianwords 3mo ago
    Why do you think the spending is ridiculous? Do you not agree that we've had a step improvement in coding, mathematics, search/retrieval?
    - dgellow 3mo ago
      It's irrelevant... The spending commitment OpenAI has for 2030 is larger than the projected AI infra market in 2030. And we still haven't seen a proof that agents contribute positively to consumers ROI. It's a capital misallocation problem, a number problem, that has nothing to do with subjective feelings regarding LLMs
      - simianwords 3mo ago
        In that case, if the numbers do work out and labs make shit load of money, would you agree you were mistaken?
        dgellow 3mo ago
        no... please actually look at the information we know about AI vendors and how much debt they have. OpenAI is projected to have more than $20B in losses for 2026. They wouldn't _just_ need to make a shit ton of money, they would need to first become profitable, then do a shit ton of money while staying profitable, for multiple years, just to match their spending commitment. That's not even taking in account their ROC and company valuation
        simianwords 3mo ago
        Sure but if they do make money and profit, your concern would have been invalid in hindsight wouldn’t it? Your concern seems to be profitability of OpenAI but that’s easily falsifiable. I’m not saying it’s 100% but I’m saying if they do turn out profitable, your concern wouldn’t have been valid.
        dgellow 3mo ago
        depends on the exact chain of events, but if the economics are indeed solved, and AI labs are found to have a sustainable business model, that concern would be addressed. I would still see the technology as anti-human and would still see the concept of an agentic economy as deeply unserious, wasteful, and risky. Not exactly sure what your point is though. FWIW my concerns aren't only for OpenAI, it's just the poster-child of the AI bubble. SpaceX AI strategy and investment is a complete joke (the S-1 they filled is an insult to a reader's intelligence). Anthropic seems to have been more cautious but has the same fundamental economic problems. Nobody in that whole industry has a moat, the top AI vendors are way too exposed to Chinese/open-weights labs. The datacenters debt investment vehicles (SPVs) look extremely shady and made to obfuscate the underlying assets. The HBM manufacturers seem to be going through one of the most violent boom-burst cycle ever. Oracle is... doing Oracle things, I would be shocked if that company is still alive in its current form in 5y. And so on

rglover 3mo ago

I'm the author. If you do go fast and don't deprive your customers of your product but the product breaks and doesn't deliver its intended value, is that a better tradeoff? Re: AI development, I'd say it's foolish to slow down research, but worthwhile for rollouts to the market. The time that should have been used for preparing people/businesses was wasted on fearmongering campaigns. Now we have a fire hose of slop that's out of control. Why? Rushing.

simianwords 3mo ago
> If you do go fast and don't deprive your customers of your product but the product breaks and doesn't deliver its intended value, is that a better tradeoff? Its not a good tradeoff if it breaks it enough that people move off your product, I obviously concede that. Ultimately a firm should be incentivised to improve their long term profits after accounting for externalities to the world. Fast or slow is an implementation detail. >Re: AI development, I'd say it's foolish to slow down research, but worthwhile for rollouts to the market. The time that should have been used for preparing people/businesses was wasted on fearmongering campaigns. Now we have a fire hose of slop that's out of control. Its not clear that your subjective view of slop is reason enough for the world to slow down on AI rollout and I'm glad I don't live in a world that doesn't necessitate slowing down due to whims and fancies of a few people's subjective aesthetic takes. Ironically for me, all the companies _do_ want to slow down their pace for reasons related to actual safety. I think you would appreciate the cause and the foresight (if any). https://www.pacingthefrontier.com/ https://www.pacingthefrontier.com/
- rglover 3mo ago
  > Its not clear that your subjective view of slop is reason enough for the world to slow down on AI rollout and I'm glad I don't live in a world that doesn't necessitate slowing down due to whims and fancies of a few people's subjective aesthetic takes. This is a misunderstanding of what I'm talking about. It's not about "slop" in and of itself, it's about what reality that slop inevitably creates. There's a lot of focus on the output ("ermagherd you wrote this with AI"), but very little on the ripple effects/second-order effects (people spinning up cults, ending marriages, and gambling away their money [1]). That's not just cute "look at the freaks" stuff, that's a serious, civilizational-level problem 1-2 decades out. You're right that my subjective opinion means dick-all beyond being that of an experienced, educated user of this stuff. But I'd at least hope that people take heed of what I'm saying and start to push back when it's objectively clear that decisions are closer on the spectrum to disaster than they are to benign mediocrity. Thank you for sharing that link, this is exactly what I'd hope to see happening behind the scenes. [1] https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chatbot-users-lives-wrecked-by-delusion https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chat...