BabaIsPissed [he/him]

  • 1 Post
  • 32 Comments
Joined 2 years ago
cake
Cake day: June 10th, 2022

help-circle
  • We consistently find across all our experiments that, across concepts, the frequency of a concept in the pretraining dataset is a strong predictor of the model’s performance on test examples containing that concept. Notably, model performance scales linearly as the concept frequency in pretraining data grows exponentially

    This reminds me of an older paper on how LLMs can’t even do basic math when examples fall outside the training distribution (note that this was GPT-J and as far as I’m aware no such analysis is possible with GPT4, I wonder why), so this phenomena is not exclusive to multimodal stuff. It’s one thing to pre-train a large capacity model on a general task that might benefit downstream tasks, but wanting these models to be general purpose is really, really silly.

    I’m of the opinion that we’re approaching a crisis in AI, we’ve hit a barrier on what current approaches are capable of achieving and no amount of data, labelers and tinkering with architectural minutiae or (god forbid) “prompt engineering” can fix that. My hopes are that with the bubble bursting the field will have to reckon with the need for algorithmic and architectural innovation, more robust standards for what constitutes a proper benchmark and reproducibility at the very least, and maybe, just maybe, extend its collective knowledge from other fields of study past 1960’s neuroscience and explore the ethical and societal implications of your work more deeply than the oftentimes tiny obligatory ethics section of a paper. That is definetly a overgeneralization, so sorry for any researchers out here <3, I’m just disillusioned with the general state of the field.

    You’re correct about the C suites though , all they needed to see was one of those stupid graphs that showed line going up, with model capacity on the x axis and performance on the y axis, and their greed did the rest.




  • got the Samsung buds pro 2 at half price recently and I kind of like them, but they were a bit underwhelming even at that price. I’ve never spent a lot on audio in general, so they were actually a big improvement, but there was no “wow” factor or anything. Plus having to install bloatware that asks for all permissions under the sun sucks (why the fuck would a settings menu want to know my location???).

    I do think you underestimate how nice the noise cancelation can be though. I moved to a big city and my hick ass cannot deal with all the fucking noise. Plus I’m clumsy and end up getting wires caught on everything, which means wire stuff also becomes e-waste fairly quickly.


  • It reads as parody so much that I was actually kind of enjoying it. I mean the scene with the Teslas crashing, Havana Syndrome, “Death to America, I remember it from the videogame”, the 13 year old girl quoting West Wing, all the nonsense pontificating about random bullshit like Friends, the camera zooming and twisting for no discernible reason, I thought I was picking up on some deep contempt for lib neurosis and vapidness and it got a laugh out of me a couple of times. Pausing it to check out who made it was a mistake, soured the rest of the thing. It might have been ruined by the runtime anyway so no big loss.



  • The video consists of like 2 hours of some examples of youtube plagiarism, with discussion of content mills and the beginning of an interesting point about how plagiarists view the people they steal from as lesser, which is not expanded upon as much as it should IMO.

    The other 2 hours are about James Somerston, a gay video essayist that basically Frakensteined a bunch (if not all) of his videos from queer authors, some well known, a lot of them not. By the end Hbomb makes a good point about erasure, and how young queer people don’t understand their history in part because of people like Somerston.

    I’m generally not annoyed by length since I’m a zoomer and watch everything at 2x or more, but in this case I get the point because it actually took me 2 real time hours to watch and I felt there was a lot that could have been cut. I still won’t watch the 3 hour scorcese movie, fuck all of you, movies ARE too long now.

    I went on a ramble about the video that should probably be a separate comment, feel free to ignore

    So regarding video length I think there’s some value in going into detail about how plagiarism takes place. Some of this context is also relevant as a way to preempt any shitty response (for example he took the time to explain that Somerston’s assistant writer is most likely not in on it and how his boss has shown signs of setting him up as a scapegoat).

    He also seems to genuinely care about James’s plagiarism because he’s bi himself. I’m currently having a bout of insomnia, and was reading The Gentrification of the Mind before making the mistake of opening hexbear and seeing a new hbomb video was out, and while these two are not comparable in content, I found it interesting to experience them back to back, and since a bit of the vibe is there, I believe Harris is sincere about why he cares about it (also Vito Russo’s name popped up in both, so I guess this is a sign that I should add The Celluloid Closet to my reading list).

    However, the video does feel petty. Hbomb has this mean streak to him that served him really well when he was directly responding to right wing talking points, but is a lot less useful when talking about stuff like this, which becomes a problem when it’s a large chunk of the video. He kind of recognizes it too, saying this feels like a drama video, and how he’s donating all ad revenue to people James plagiarized from.

    It does feel more appropriate when you consider that most of these people are reactionary pieces of shit and that should have been a much larger part of the video. He mentions it a bit (the first guy is a chud, internet historian tries to hide that he’s a chud, Somerston came from business school and seems to hate women) and talks about contempt for the people they copy from, but I feel there’s a lot more to dig into. What about contempt for the audience? What is the frame of mind of people that trend chase for years, sometimes decades, in order to garner an audience? That think regurtitating Wikipedia is worthy of other people’s time? He says it was always like this mentioning AVGN copycats, but was it really? While the incentive structure didn’t change how plagiarism takes place, didn’t the kind of people that did the plagiarizing change? I think exploring this thoroughly is a lot more interesting than “showing the receipts” by comparing the copied work to the original for most of the runtime.

    I still think it was worth a watch, but that’s because I was already familiar with Somerston and some of the other people and they gave me the video essay equivalent of the ick. This should have been 2 hours at most.






  • They do once their depression gets better though? Anhedonia, loss of interest/libido/attention/whatever the fuck else are symptoms of depression. I’m all for self-improvement, my own mental health improved greatly as a result of trying to improve myself, to the point I consider myself no longer depressed. But we’re social creatures and no one builds self-confidence and mental resilience in a vacuum. It’s often up to the depressed person to put themselves out in situations where this can happen, but sometimes it does not work out for whatever reason and the whole thing is a long process. In this situation self-compassion is a lot better than telling yourself you’re a sack of shit.

    Also, isn’t the interesting life thing all backwards? If you like a person you get curious and find them interesting. If I like a guy I’ll find what they are into cool, be it singing, playing chess or knowing a lot about bugs.

    No one is owed that kind of attention, but most people are worthy of compassion.