Hello thanks for visiting my profile.

For any picture posts I make with the [OC] tag, I provide a license for you to use my photo under the terms of CC-BY-SA-4.0. You may DM me for questions.

  • 41 Posts
  • 3.32K Comments
Joined 3 years ago
cake
Cake day: March 14th, 2023

help-circle




  • AI can be a useful tool no doubt and has cool and harmful applications. But let’s not be delusional about it and anthropomorphize AI. It doesn’t think. It doesn’t go rogue. It’s a non-deterministic phrase guessing algorithm. I’ve tried chatbots, thing generators, had a friend try coding tools, ran my own models, and I find parts of it impressive and useful in some respects, but at this moment, I don’t really buy the narrative that it’s going to change everything or fix the limitations I see with transformers-based AI models. The best description of it I heard was a lossy compression algorithm, it can take a lot of knowledge and compress it but there are artifacts, hallucinations and weird stuff sometimes coming out the other end.

    I’m not against AI itself per se, but I’m very much against the bubble hype-cycle, techbros being locked into a strange belief that pumping money into this thing forever is somehow going to go take over the world on its own volition. On the contrary, my understanding is corporate greed and capitalism is already accomplishing that - firing the workers, concentrating wealth and controlling the world - no matter how much AI is cited as the reason for any of that.


  • We have mayors and folks courting GEO group to run private prison and detention camps in their town…

    They know it’s depressing and immoral but they rationalize it to themselves because they know it brings lots of high paying jobs with low educational requirements and activity to their region.

    Data centers on the other hand are only take and no give. There aren’t many local staff needed, it doesn’t require that much cleaning and security that don’t pay that well, the one or two high paying jobs are specialized and shipped from another state probably, the construction companies are likely to be mostly from outside the county, and the center will be a draw on local resources to serve AI workloads everywhere else in the world.












  • A Project 2029 cannot be a collection of Democratic Party agenda items. It must articulate a broad new conception of the nation’s political order — one that will guide the way a future Democratic-led government might wield power. Above all, Democrats must have a plan for reconstruction — for building something new on the wreckage of what President Trump, MAGA and the Republican Party have wrought — not restoration of what was.

    The whole article puts eloquent prose to the thoughts in my head (looking from outside the US) as to what the USA needs to recover post-Trump; it’s more than striving to going back to normal.

    The issue is that centrist Democratic voices have been kind of developing their vision by committee over any aims to make it cohesive and comprehensive. In my view the plan has to recognize and tackle these three things:

    • Internal complacency within the Democratic party to do good things for people rather than just talk about them
    • Reforming the outdated systems that simply relied on trust that the system wouldn’t corrupt itself or overlook corruption.
    • Fixing the destruction caused by the Trump administration, then taking steps to prevent a cult like Congress and a bought-and-paid-for judiciary to surrender the republic to the executive.

    The important thing is you can’t grandfather in the corruption while you fix it.



  • To clarify, AI companies charging the cost that would make the inference profitable for them, against the operating costs and financing costs on new capital expenditures (new data centres, new compute and new model training*), is more than what most people appear to be willing to pay. That cost is indeed more than just the cost of inference incurred by the AI company.

    *(I’m being generous and including model training as capex for the sake of argument, even if I personally think to continue the hypetrain, continuous model improvements are core to AI companies’ operation.)