hraegsvelmir

@hraegsvelmir@ani.social

• ⁨Comment⁩ on ⁨Steam's AI disclosures are akin to "forcing developers to list their voting history", says man who must hate seeing eggs virtue signal that they're free-range⁩ ⁨⁨2⁩ ⁨weeks⁩ ago⁩:

  Well people can’t even agree on who won the last election or even if the world is round, so I can only speak to my point of view on things.

  I don’t think these are the best examples to use for your case here, giving both touch on matters of fact that are only questioned by fringe conspiracy theorists that the rest of the world regards as absolutely crazy, without any basis in reality. Yes, people do passionately believe them, but that doesn’t make them anyore true. It’s like trotting out the conservative chestnuts that evolution is just a theory, or that anthropogenic climate change hasn’t been the consensus position amongst scientists and framing it as a quack theory that some evil, socialist scientists have politicized because they hate US oil barons making money.

  That aside, any serious discussion of AI’s inevitability needs to involve a feasible roadmap to making it sustainable at a material level. I won’t fully discard the possibility, as there’s no telling what the future may hold, but I haven’t seen any indication that there is a viable way to make these systems run with performance comparable to what they currently offer, without requiring several planets’ worth of water, hardware and power capacity. They can only do their “move fast and break things” schtick for so long before people get sick of it, and you’re already seeing people sour on data centers, as well as AI more generally.

  The idea that AI will always be here, in something approaching its current state, now that we’ve invented it is just groundless, in my opinion. Yes, local LLMs will continue to exist, but they will have trade offs that I don’t see being palatable for the broader market. Using AI to make a quick meme based on a prompt doesn’t have the same draw when it takes 40 minutes to finish what the current money pit versions of AI will do in seconds.

  I think it’s fair to assume AI will be around in some capacity, but will likely be far different from what we currently have, and restricted to domains where it actually makes sense to use, like some medical applications for analyzing patient imaging to suggests issues that humans might overlook. As companies are currently learning, though, the math just doesn’t work out on trying to replace human workers wholesale, and this is while the price still doesn’t accurately reflect the cost of running these models as services.

  Given all this, plus your remarks on the ethical side of the training material, it just doesn’t make any sense to discuss the inevitability of AI existing in some capacity in the future as a reason not to offer blanket opposition to the tech as it currently exists. None of these companies can currently turn a profit just on the basis of their hardware and energy costs, so how likely is it they could offer reasonable compensation for training materials without this price becoming even more unreasonable.

  I think these drastically reducing these costs and ironing out the wrinkles should have been done before they ever rolled these systems out to the general public. Now, they’re up against a wall, trying to find a way to do so before the backlash becomes so severe, they can no longer find steady streams of investors willing to sink money in an increasingly toxic product.

• ⁨Comment⁩ on ⁨⁩ ⁨⁨4⁩ ⁨weeks⁩ ago⁩:

  Is it really shooting itself in the foot, though? There are plenty of FOSS projects that companies benefit from and never contribute back to, either monetarily, or in the form of writing code. They provide an easy option to make it look more professional while helping to support the project, and you could always fork it and do what you need to with your fork. The idea that developers should do a thankless job to maintain their software in a way that complies with random corporate policies in order to be more widely adopted strikes me as setting up devs to burn out and crash from having all the joy ground out of their projects for them.

  UNESCO, the Wikimedia foundation and other major organizations have already implemented it and gotten onboard, so I would suggest your coworkers are being a bit foolish in completely disregarding it because of a single mascot image that has a clear option to replace. The Dev is just saving themselves aggravation from people that were going to make endless demands of them, but most likely balk at actually supporting the project to enable features that only matter to their organization, but not to the Dev themselves or others.

• ⁨Comment⁩ on ⁨Steam Frame Release⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

  Sure, but with production stuff at scale like that, even if contracts fall through, it doesn’t necessarily mean they’ll be able to switch right over to running consumer RAM down the lines. Unless the AI bubble bursts hard and soon enough for them not to have contracts to fulfill by March, I wouldn’t count on them being able to get meaningful production going during Q2. Switching over from producing one SKU to another can be a crazily time intensive problem on its own, and that’s leaving aside whether they have enough raw materials on hand to keep the lines busy producing those products. If they have some stupid Just in Time inventory system in place, we’ll get to follow their supply lines as we see where new logistical bottlenecks are emerging in the process.

• ⁨Comment⁩ on ⁨Steam Frame Release⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

  but I fear we have at least till the end of this year, possible Q2 of next year to see it happen.

  Considering at the start of August, I was already seeing stories that all the fabrication capacity for RAM has been completely spoken for and booked through the end of 2027, I would expect it to keep going quite a bit longer still.