4 ms·

Good point, dev is a bad word. "beta" or "early access" might have been a better term. I don't really believe that this is causing harm, however. GenAI is stil

Good point, dev is a bad word. "beta" or "early access" might have been a better term. I don't really believe that this is causing harm, however. GenAI is still so new, and Gemini in particular, that I don't think anybody is genuinely basing their political opinions off its output. If this product had achieved the cultural legacy of ex. Wikipedia I would understand the concern

MBCook 3y ago
Yeah the “if it causes harm” thing is certainly debatable in this situation. i was commenting more generally. I’m tired of the “its beta give them a pass” attitude I see used in many situations even if the harm is far more clear. I’ve seen Autopilot/FSD/Cruise defended that way. “Oh if it runs over someone it’s beta so it’s ok” makes no sense to me, but that’s basically the attitude I’ve seen but not stated that cleanly. I don’t really believe in the idea of a public beta period. You’re just making a rough incomplete release. I understand saying “we put it out early, sorry there are a few bugs and it looks rough”. But there has to be a line somewhere. And to me harm is on the other side of that line.
- tech_ken 3y ago
  >I understand saying “we put it out early, sorry there are a few bugs and it looks rough”. But there has to be a line somewhere. And to me harm is on the other side of that line. Very true, I'm definitely unenthused about the rush to market with all these hype-y AI products. At a minimum it seems fair to criticize Google for releasing something this half-baked, instead of taking the time to get it right and trusting that the product's value will make up for missing the feeding frenzy. I just feel like this is the consequence of bad product thinking, rather than the larger cultural agenda (seemingly) imputed by the OP's author. If it does lead to unambiguous downstream harm I certainly won't defend that, I just feel that the product space is innocuous enough (at least relative to the ex. AV space) that I feel like this is more of a cultural war issue than an actual technology concern.
  - MBCook 3y ago
    I don’t think Ben was trying to say it’s a cultural agenda thing. It’s clear to me he wants to avoid that third rail for his point. I think he spelled it out well on his subscription only podcast Dithering with John Gruber. They were afraid of being criticized for only white men when you asked for a picture of a CEO (likely based on input data) or a woman for a nurse or some other traditionally stereotypical profession. So they seem to have put in something to ensure diversity to try to correct for bias in the input data. That seems good to me. But they overcorrected in a way it just wouldn’t show you white people. Turns out the Nazis and founding fathers were more diverse than I was taught. Oops. But worse, and the big problem Ben pointed out, was there was no way this wasn’t known internally. And the company culture seems to be at a point where either you can’t speak up or it’s ignored because diversity is good even if it’s at the cost of all accuracy on easily verifiable facts. And that’s a problem. They avoided the culture war so hard out of fear they crammed themselves back into the culture war and it blew up in their face. It was an unforced error that they must have seen coming. So this whole incident seems very indicative of a culture/management problem inside Google.
    - tech_ken 3y ago
      Hm okay fair enough, I'll check out the podcast I do think that's a better format for communicating some of the details of this type of criticism. > It was an unforced error that they must have seen coming. To some extent this feels like something that's happening a lot in tech across the board, not just in 'culture war' type issues but across several other fronts (the various Metaverse flops, for example). I do think this indicates an overabundance of groupthink, but part of me wonders if its just a consequence of these companies becoming so big and working on so many fronts that it's basically impossible for them to avoid stepping on every rake. Still, they did seem like they genuinely wanted this to take off so concerning that couldn't be prepared for even this basic failure mode (which is already a pretty well-trod problem space).
      - MBCook 3y ago
        I agree. A lot of it just seems to be general big company ossification and trouble avoidance. Don’t do anything anyone might criticize you for = don’t do much of anything innovative. But AI is very innovative. Since it’s trained in biased/imperfect data and just hallucinates in general they seem to be in a minefield now by mistake.

MBCook 3y ago

Yeah the “if it causes harm” thing is certainly debatable in this situation. i was commenting more generally. I’m tired of the “its beta give them a pass” attitude I see used in many situations even if the harm is far more clear. I’ve seen Autopilot/FSD/Cruise defended that way. “Oh if it runs over someone it’s beta so it’s ok” makes no sense to me, but that’s basically the attitude I’ve seen but not stated that cleanly. I don’t really believe in the idea of a public beta period. You’re just making a rough incomplete release. I understand saying “we put it out early, sorry there are a few bugs and it looks rough”. But there has to be a line somewhere. And to me harm is on the other side of that line.

tech_ken 3y ago
>I understand saying “we put it out early, sorry there are a few bugs and it looks rough”. But there has to be a line somewhere. And to me harm is on the other side of that line. Very true, I'm definitely unenthused about the rush to market with all these hype-y AI products. At a minimum it seems fair to criticize Google for releasing something this half-baked, instead of taking the time to get it right and trusting that the product's value will make up for missing the feeding frenzy. I just feel like this is the consequence of bad product thinking, rather than the larger cultural agenda (seemingly) imputed by the OP's author. If it does lead to unambiguous downstream harm I certainly won't defend that, I just feel that the product space is innocuous enough (at least relative to the ex. AV space) that I feel like this is more of a cultural war issue than an actual technology concern.
- MBCook 3y ago
  I don’t think Ben was trying to say it’s a cultural agenda thing. It’s clear to me he wants to avoid that third rail for his point. I think he spelled it out well on his subscription only podcast Dithering with John Gruber. They were afraid of being criticized for only white men when you asked for a picture of a CEO (likely based on input data) or a woman for a nurse or some other traditionally stereotypical profession. So they seem to have put in something to ensure diversity to try to correct for bias in the input data. That seems good to me. But they overcorrected in a way it just wouldn’t show you white people. Turns out the Nazis and founding fathers were more diverse than I was taught. Oops. But worse, and the big problem Ben pointed out, was there was no way this wasn’t known internally. And the company culture seems to be at a point where either you can’t speak up or it’s ignored because diversity is good even if it’s at the cost of all accuracy on easily verifiable facts. And that’s a problem. They avoided the culture war so hard out of fear they crammed themselves back into the culture war and it blew up in their face. It was an unforced error that they must have seen coming. So this whole incident seems very indicative of a culture/management problem inside Google.
  - tech_ken 3y ago
    Hm okay fair enough, I'll check out the podcast I do think that's a better format for communicating some of the details of this type of criticism. > It was an unforced error that they must have seen coming. To some extent this feels like something that's happening a lot in tech across the board, not just in 'culture war' type issues but across several other fronts (the various Metaverse flops, for example). I do think this indicates an overabundance of groupthink, but part of me wonders if its just a consequence of these companies becoming so big and working on so many fronts that it's basically impossible for them to avoid stepping on every rake. Still, they did seem like they genuinely wanted this to take off so concerning that couldn't be prepared for even this basic failure mode (which is already a pretty well-trod problem space).
    - MBCook 3y ago
      I agree. A lot of it just seems to be general big company ossification and trouble avoidance. Don’t do anything anyone might criticize you for = don’t do much of anything innovative. But AI is very innovative. Since it’s trained in biased/imperfect data and just hallucinates in general they seem to be in a minefield now by mistake.