• 0 Posts
  • 230 Comments
Joined 1 year ago
cake
Cake day: July 7th, 2023

help-circle



  • maniclucky@lemmy.worldtoMemes@lemmy.mlAI bros
    link
    fedilink
    arrow-up
    6
    ·
    2 months ago

    Absolutely. It’s why asking it for facts is inherently bad. It can’t retain information, it is trained to give output shaped like an answer. It’s pretty good at things that don’t have a specific answer (I’ll never write another cover letter thank blob).

    Now, if someone were to have the good sense to have some kind of lookup to inject correct information between the prompt and the output, we’d be cooking with gas. But that’s really human labor intensive and all the tech bros are trying to avoid that.


  • maniclucky@lemmy.worldtoMemes@lemmy.mlAI bros
    link
    fedilink
    arrow-up
    8
    arrow-down
    3
    ·
    2 months ago

    Gradient descent is a common algorithm in machine learning (AI* is a subset of machine learning algorithms). It refers to using math to determine how wrong an answer is in a particular direction and adjusting the algorithm to be less wrong using that information.







  • I’ve no significant opinion of India beyond anti-Modi, and that’s a product of John Oliver. Most of my engineering team are Indian and some I like, some I tolerate. And a fear of Indian traffic by reputation alone.

    But you could swap “American” with “Indian” in that first paragraph, change nothing else, and it be largely (if not entirely) accurate.




  • maniclucky@lemmy.worldtoScience Memes@mander.xyzHero
    link
    fedilink
    English
    arrow-up
    5
    ·
    5 months ago

    My problem with your example is that the loner didn’t have comparable value. If it was supporting other things, then it failed. If it was doing something non obvious, it shouldn’t be compared to the support. It feels fallacious, though I can’t name one specifically.

    System sight is itself an issue. Many companies evaluate an employee solely on some performance metric, typically tied to money. Because it’s easy (and lazy).

    I’ve had several positions where my task was to keep things running. I added no value, I prevented loss. And those positions get screwed because they’re very difficult to quantify worth and very hard to see (and if it doesn’t create money, they don’t care). You only notice them when something goes wrong. Such an employee may keep everything running all year and get a “meets expectations” because there’s an upper limit on how much contribution the system sees, and the system doesn’t want to put in the effort to see better. I may have had to climb over an air handler to get to a transducer to calibrate, but that’s not sexy and even if I report such effort, it’s what I’m supposed to do (even if I wasn’t, weekend nights are weird).

    No one is going to write down “keep machine running 80% of the time” because people unassociated with the task will insist that 100% is the expectation, despite that being unreasonable.

    A system built of people is not a black box. We can see them and evaluate them based on the task they’re supposed to do, but the evaluators don’t want to put in the effort to do their tasks in a way that means more work for them.

    There’s a comment to be made also about scope creep for a position so that a company doesn’t have to hire marketing and engineering if they can get the engineers to do it. Despite them being suboptimal for the task. Something something down with unrestrained capitalism.

    Ok. I’ve lost the plot at this point and made my point. Have a good one.


  • maniclucky@lemmy.worldtoScience Memes@mander.xyzHero
    link
    fedilink
    English
    arrow-up
    11
    arrow-down
    1
    ·
    5 months ago

    That’s a pretty contrived setup. If the two top components are not factored into the performance of the whole and they are both defined by their ability to improve other components, then the one doing it’s own thing is not, in fact, a top performer. It’s task is to support others and it fails to do so.

    And what if the loner’s task is foundational? It doesn’t have much direct output, but if he’s gone and everything else goes to shit? Those ones are very hard to measure. I know, that’s been my job for a good portion of my career. And things like that are common. Expecting a given performer, say an engineer, to also be good at public speaking has always struck me as impractical.






  • And we diverge again, though not hugely so.

    I feel that you’re unnecessarily blaming statistics (which as someone who does them, doing them well takes work. Though no shortage of people doing them badly, I digress) for a different societal ill: mob mentality.

    The ideal solution is to investigate each instance of rape and mete out justice appropriately. Obviously that’s not going to happen. And the current state of affairs is also no good. Obviously, there isn’t a legal way to really handle any of it because everything we’ve mentioned is a crime. It kinda comes down to a cultural shift. People need to be be more willing to accept that rape occurred (because fears of not being believed are pretty valid sadly) and also that justice takes too much time (also a big social problem) and that there should be a lot more stigma about false reporting and a whole bunch of other things. I’m not gonna solve this in a lemmy comment, but I’d hazard that we all need to listen to each other (myself included) to start. I still contend the reason we’re having this conversation is that not enough people listen to anyone that does get raped in addition to a system that hasn’t caught up to the population or the times. I further hazard it isn’t that people are unaware of the horror of being falsely accused, just that it isn’t the biggest issue at hand (though that is a bitter statement for the victim).

    There’s no good easy solution, but progress can be made.