Aerrol's recent activity

  1. Comment on The Funny Men in ~creative

    Aerrol
    Link
    Wow, really enjoyed this. Thanks for sharing. The ending really landed for me. (Sorry I don't normally comment on poetry so I'm not sure how to respond properly)

    Wow, really enjoyed this. Thanks for sharing. The ending really landed for me. (Sorry I don't normally comment on poetry so I'm not sure how to respond properly)

    4 votes

  2. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    Link Parent
    That's true. I'm not confident where the line between actionable and dreaming too small lies in this case. Probably in an advisory non profit that builds global consensus on definitions and guard...

    That's true. I'm not confident where the line between actionable and dreaming too small lies in this case. Probably in an advisory non profit that builds global consensus on definitions and guard rails? Get experts and academics involved.

    2 votes

  3. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    Link Parent
    Uh, okay? And I'd say over half of the people I know and work with are sick of AI hype and worried about the many problems currently being exploded by these companies. Nevermind broader...

    Uh, okay? And I'd say over half of the people I know and work with are sick of AI hype and worried about the many problems currently being exploded by these companies. Nevermind broader perspectives from experts, academics, and actual facts like skyrocketing prices and power consumption.

    Also very strange to use another website to describe users of this one.

    9 votes

  4. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    Link Parent
    I agree that it feels unlikely right now, but I strongly believe in the power of optimism and pushing for positive change. We shouldn't accept it not happening just because the Governments right...

    I agree that it feels unlikely right now, but I strongly believe in the power of optimism and pushing for positive change. We shouldn't accept it not happening just because the Governments right now suck. We should be pushing for better Governance.

    4 votes

  5. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    (edited )
    Link Parent
    Fair point about civility - I will try and keep it respectful. But my response is that jumping from "we need a government regulatory body" to "this is either a total surveillance state or...

    Fair point about civility - I will try and keep it respectful. But my response is that jumping from "we need a government regulatory body" to "this is either a total surveillance state or pre-computer society" is a wild jump that is highly disrespectful and smells a lot to me of sea lion-ing. So that's where my strong response came from. I think this subsequent response is much more civil.

    To go back to your response - the link quoted notes that they received a warning way back in 2019 and failed to comply with said warning even 3 years later before they were shut down. That's incredibly reasonable to me and frankly I'd probably prefer they be shut down much before then. As for months of work and millions in costs to respond to an audit? Those are all things that can and should be tinkered with as we try and find a right-sized response, rather than throwing the whole concept of government oversight out the window.

    Frontier research in AI (or, more specifically, LLMs) is almost entirely being driven by the large corporations at this point. Roll-out as I reference it has nothing to do with sharing your findings. Roll-out as I mean it is releasing a commercial product for broad consumption and being backed by enough compute to deploy at scale. So a multi-billion dollar initiative at minimum. The core companies and behaviour we need to be targetting could be selected by a combined threshold of: product deployment (how many people have access to the tool?) + market cap + proposed capabilities (a nice by-product here would be a cooling of the constant spam of OUR AI IS OMNISCIENT AND SOLVES EVERY PROBLEM).

    Thinking a bit deeper about it, I see no reason why we shouldn't have some social controls on AI research more broadly. We have controls over human genetic research, nuclear power research, and CERN is highly regulated, to give a few examples. Why is something that is globe-spanning and society-impacting like AI excluded from this?

    9 votes

  6. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    (edited )
    Link Parent
    I agree with a lot, but not all, of what @SloMoMonday says below. Most importantly, I agree with the main point that this isn't an optimal economic strategy even from a cold geo-political conflict...

    I agree with a lot, but not all, of what @SloMoMonday says below. Most importantly, I agree with the main point that this isn't an optimal economic strategy even from a cold geo-political conflict perspective. That said, it doesn't actually address your question directly, so I'll respond too.

    I'll focus on your actual question in 2 parts.

    First, an ideal but unlikely outcome: We manage to negotiate in good faith with the Chinese Government to come to agreed upon global standards of research and care. We have similar international alignment already on: the Law of the Sea, Nuclear Arms Control (ish), and intellectual protection law (I know the Chinese are sketchy on enforcement but they are party to the treaty). So, in this ideal world each country has their own governing body, but those bodies are largely aligned, allowing us to more controlled manage this period of flux.

    Realistically, I would say at minimum we do what the Chinese are already doing: none of the Chinese companies are developing things in a way which the Communist Party isn't already approving. They're already enforcing stricter controls on the usage of personal information for instance (lol of course this doesn't apply to the Government's uses of AI but that's another topic). So you can still encourage pretty break-neck development while also putting some government back stops and "final say" authority on how this research is being conducted.

    6 votes

  7. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    Link Parent
    Wow, talk about a "slippery slope" strawman. The FDA can shut down your business, but how often do they do that? Do you have any evidence that this is some kind of foundational problem? We are...

    Wow, talk about a "slippery slope" strawman. The FDA can shut down your business, but how often do they do that? Do you have any evidence that this is some kind of foundational problem? We are constantly seeing medical companies and drug companies growing and delivering new solutions, so I'd hardly call it a police state solution. Are you saying that we should abolish the FDA?

    As for the actual topic at hand, it absolutely is a matter of size and scale. The Government is a living entity, not a permanent, unchanging monolith. We deal with large players who are using entire States' worth of power, not random people tinkering with open source AI tools at home. If one day we reach a point where every AI model is small enough to be distributed at that scale, we've clearly reached a new paradigm shift (again), and should be reflecting on the possibilities for governance at that point.

    10 votes

  8. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    (edited )
    Link Parent
    I think my ideal (and I recognize the unfeasibility here due to cost and complexity) would be a highly competent regulatory authority that puts out regularly adjusted limits on: Data centre...

    I think my ideal (and I recognize the unfeasibility here due to cost and complexity) would be a highly competent regulatory authority that puts out regularly adjusted limits on:

    1. Data centre rollout
    2. Emissions (tied to 1) and other power generation
    3. Data centre utilization by a single entity
    4. Frontier research being conducted, period
    5. Frontier research being rolled out commercially

    It's insane that 4 and 5 in particular are being researched and released with ZERO CONTROLS. Look at biomedical research - we have a highly structured, rigorous, and regimented process for the release of drugs (or we did before the MAGA disease struck). We should, hypothetically, be able to have an AI equivalent.

    14 votes

  9. Comment on The AI disruption has arrived, and it sure is fun (gifted link) in ~tech

    Aerrol
    Link Parent
    This drives me nuts. Yes we can stop. That is the purpose of the Government - it just sucks at being proactive. We should be legislating and non-profiting real guard rails and thinking deeply...

    The market keeps convulsing, and I wish we could hit the brakes. But we live in a brakeless era.

    No matter where you work, my hunch is this is coming for you. Have you noticed the software you use every day adding “A.I. features”? That’s the top of the slippery slope. Whatever unifying principle equates to ship risk in your industry, people are trying to mitigate it with A.I. Insurance, finance, architecture, manufacturing, textiles, every kind of project management — they want to automate it all through A.I.

    This drives me nuts. Yes we can stop. That is the purpose of the Government - it just sucks at being proactive. We should be legislating and non-profiting real guard rails and thinking deeply about what we want to happen. Instead we've let the AI companies push us into some kind of odd determinism that this is our own only possible future.

    26 votes

  10. Comment on Drinking two-three cups of coffee a day tied to lower dementia risk in ~health.mental

    Aerrol
    Link
    Nice, gonna ride the confirmation bias and pat myself on the back for staving off dementia with my addiction

    Nice, gonna ride the confirmation bias and pat myself on the back for staving off dementia with my addiction

    35 votes

  11. Comment on Jeff Bezos bought The Washington Post to save it. Instead, with a mass layoff, he’s forced it into severe decline. in ~news

    Aerrol
    (edited )
    Link
    The positive impression Bezos left on staff in the early years is really interesting. I remember viewing his purchase incredibly skeptically, so the current trajectory seems like it was always the...

    The positive impression Bezos left on staff in the early years is really interesting. I remember viewing his purchase incredibly skeptically, so the current trajectory seems like it was always the likely conclusion. I'd love to somehow see things from Bezos' perspective, to see what changed for him. Maybe it's just billionaire brain rot over time?

    7 votes

  12. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    Link Parent
    Fixed! Thanks. I always mess those up, especially on mobile.

    Fixed! Thanks. I always mess those up, especially on mobile.

    1 vote

  13. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    Link Parent
    Yes for sure, I was trying to be specific about only positives for this but definitely agree. For any space option to be truly considered environmentally friendly, launch emissions and burn ups...

    Yes for sure, I was trying to be specific about only positives for this but definitely agree. For any space option to be truly considered environmentally friendly, launch emissions and burn ups should be considered in the equation.

    1 vote

  14. Comment on Elon Musk says SpaceX will prioritize a city on the moon instead of a colony on Mars in ~space

    Aerrol
    Link Parent
    Because he paints a very convenient/desirable sci-fi image of the future, I think. And he has a ton of money, which we somehow still equate with universal competence for some reason. Personally,...

    Because he paints a very convenient/desirable sci-fi image of the future, I think. And he has a ton of money, which we somehow still equate with universal competence for some reason.

    Personally, I'm just relieved that we're likely now to see Artemis/Lunar funding maintained instead of abandoned in some icarus-like insane rush to put people on Mars to die within the decade.

    11 votes

  15. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    (edited )
    Link Parent
    I'll just paste my comment from elsewhere in this thread - basically the issue is that's a harder sell. Satellites beaming laser-power to the surface sounds more intuitively challenging than...

    I'll just paste my comment from elsewhere in this thread - basically the issue is that's a harder sell. Satellites beaming laser-power to the surface sounds more intuitively challenging than cooling in very cold (but awful heat transmissivity) space.

    There are a lot of well-funded space based solar power/power beaming initiatives going on as well, though.

    2 votes

  16. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    Link Parent
    To take off my rage hat and put on my optimistic futurist hat, here are some potential benefits of space-based data centres (assuming you somehow solved cooling): Effectively infinite solar power,...

    To take off my rage hat and put on my optimistic futurist hat, here are some potential benefits of space-based data centres (assuming you somehow solved cooling):

    1. Effectively infinite solar power, so you can run forever with low cost, hypothetically eventually covering the heavy cost of launch and deployment through those savings.
    2. No environmental damage on Earth from the heat emissions and real estate being taken up.
    3. Security from weather and other turbulent events, as you only have to worry the solar wind (oh hey that's actually pretty major the further out from Earth you get but I'm being positive here...)
    4. The one I'm most interested in: deployment of edge-computing solutions. Rather than massive data centres, I think there's actually a lot of very, very interesting use cases you could have from making a space "cloud" centre that provides compute power to remote assets on the Moon and farther. People forget that even the Moon has notable delays - a quick google got me 2.7s to send a radio or light message to and from the Moon. This problem gets much worse when you start to take into account things like:
    • Line of sight (what do you do when the Moon's blocking your comms point with Earth?)
    • Bandwidth (we actually can't just beam terabytes of data back to Earth easily - and that could be a huge problem for deploying most AI/ML systems in space)

    So with an edge-computing solution - call it a data centre orbiting the Moon - you could enable a lot more autonomous analysis and navigation solutions on the Moon and further. You could also parse space-telescope data coming in before it hits the choke points like the Deep Space Network which is already at capacity. Anything we can do to try and make it easier to send deep-space data back home will be a huge improvemnet.

    So! Not all bad news here. And as mentioned elsewhere, if they can solve the cooling problem, that has massive, massive potential benefits on Earth and for other space applications.

    2 votes

  17. Comment on Elon Musk says SpaceX will prioritize a city on the moon instead of a colony on Mars in ~space

    Aerrol
    (edited )
    Link
    To go the other way on my ranting about the stupidity of Elon Musk's space data centre pitch: hey, he's finally admitting that his plan to ignore the Moon and go right to Mars was monumentally...

    To go the other way on my ranting about the stupidity of Elon Musk's space data centre pitch: hey, he's finally admitting that his plan to ignore the Moon and go right to Mars was monumentally stupid after all. Nice.

    Archive Link: https://archive.is/LquIn

    8 votes

  19. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    Link Parent
    Agree completely with this. Small benefit is at least this is going to involve a larger set of interesting problems to solve than just "more data centre and GPUs pls".

    Agree completely with this. Small benefit is at least this is going to involve a larger set of interesting problems to solve than just "more data centre and GPUs pls".

    2 votes

  20. Comment on Voyager Technologies CEO says space data center cooling problem still needs to be solved in ~space

    Aerrol
    (edited )
    Link Parent
    As someone who's spent years working in various tech start-up circles and currently works in "new space", let me tell you: you have far too much faith in investor rigor. The whole VC-game has...

    As someone who's spent years working in various tech start-up circles and currently works in "new space", let me tell you: you have far too much faith in investor rigor. The whole VC-game has absolutely nothing to do with real results or good engineering. It has everything to do with selling hype-cycles, building interest quickly, and exiting before the bubble pops.

    A normal VC-round, even in the Space sector, aims to exit with at least 10x return on their investment within 5 years. @papasquat is mostly right, except I'd rephrase it that they don't think their investors are idiots. They think their investors are ethic-less sharks who are hunting for the next big story that they can push into the broader media and government to generate more hype. That's what this is: a lot of soul-less but smart hype-master CEOs seeing the writing on the wall for covering the planet in ever more data centres (see: increasing protests, balking at costs, trillions in promised efficiency gains not arising), and are looking for a way to pivot so they can keep the government and institutional investor gravy-train rolling without admitting that they might have a fundamental problem.

    And yes, a side-benefit of this is that a good amount of the work being done by honest engineers and scientists will hopefully be released into the broader ecosystem and that'd be laudable. But the increasing disconnect from reality in these pitches drives me insane.

    7 votes