Comments by "Taxtro" (@MrCmon113) on "Fireship" channel.

  1. Someone has to pick up garbage, clean toilets, carry things on construction sites, etc. Someone has to do the stuff that machine learning agents are telling us to do.
    43
  2. Doesn't help you understand an uncommentated, confusing python codebase where any variable can be anything and defined anywhere. Literally spend a week of my thesis with trying to figure out what's what.
    27
  3. It's actually dystopian to witness ChatGPT turn into an absolute moron if you ask it about anything illegal.
    26
  4. By that line of reasoning you should start with FORTRAN, not C. Or better yet, don't start with a high level language at all, but some early assembly language from before 1950.
    6
  5. >its trained on bad data, and it really isn't AI in my opinion That doesn't even mean anything. If you don't have information on your problem, you don't have information on your problem. Then no procedure whatsoever is better than chance. >and its only as good as the model that's been made and the data its trained on That's trivial and applies to everyone. >You ask it to solve a problem that we as humans haven't solved yet, its not going to come up with an answer. That's completely wrong. It does come up with answers of it's own. In fact it only comes up with novel answers. >If you don't work in the field of AI, try not to get caught up in the hype Anyone with some actual understanding is hyped. I'm continuously blown away by the achievements in deep learning. >The amount of basic coding errors I've seen GPT make is pretty substantial. So what? The fact that it can generate any code upon being prompted with natural language alone is incredibly impressive. You'd never be able to make anything like that. And it does that while also being able to do a million completely other things. That's absurdly impressive. You're the kind of idiot, who never admits to any achievements in AI ever, no matter what. It's an endless cycle of "oh but it can't do X. well it can do X, but X isn't impressive, it can't do Y. Y was never hard, it can't do Z..."
    4
  6.  @igorswies5913  The dude, who hates technology and modernity, and thinks he's a badass, because he uses slightly less popular technology.
    3
  7.  @alexeynezhdanov2362  That you don't need to know what happens "under the hood" is kind of the entire point of a programming language. Ultimately a programming language is a set of instructions that have a guaranteed behavior, no matter how that behavior is implemented.
    3
  8.  @TheVirtualArena24  >I'm doing data science in cse. Should I also learn like this? No. You should try to get more comfortable using higher level constructs rather than re-inventing the wheel. Ie operate on tensors with libraries like pytorch instead of using nested loops, get comfortable handling data with pandas, learn how to quickly get sth done with R, etc. How this stuff works under the hood is pretty much irrelevant.
    3
  9.  @ЏонМастерман  What's more complex? Curly braces? The actual problem is what's supposed to be complex and the programming language should make it as easy as possible. I don't know how often you've though about solving a problem in a vectorized fashion, applying operations to some of the 5 dimensions of a tensor, but I find that a lot more intellectually satisfying than fixing bugs in a C++ code.
    3
  10. Wait a minute. So the world in which Hello World wasn't printed still exists and is still as real as ours?!
    3
  11.  @sdjhgfkshfswdfhskljh3360  What corporations? We're talking about governments. Governments can shut down / take over computing centers.
    2
  12. Imo doesn't matter what they choose as long as it's object oriented.
    2
  13. That's a terribly idiotic misconception on par with the idea that a car can't be faster than an engineer.
    2
  14.  @helix8847  I can get more smart answers out of ChatGPT than I will ever get out of you.
    2
  15. And the alternative is what? Hospitals, the garbage collection, fire departments, etc aren't open source either, but you're kinda forced to use them. Nvidia has got us all by the balls. Your balls are firmly placed in Nvidia's hands. God speed your efforts to come up with a freedom alternative.
    2
  16. I had a horrible time debugging both Python and C++ code, but for completely different reasons.
    2
  17. Lots of ML stuff doesn't have good support on windows. Probably good idea just to run an Ubuntu VM if you plan to do much locally.
    2
  18.  @TheVirtualArena24  No and how "programming works" depends on what you want to do. Writing super concise haskell code for some complex algorithm, coming up with gigantic class diagrams for industry code in java, doing a bunch of experiments and hyperparameter tuning in parallel in julia - those are all different things.
    2
  19. They taught me just enough programming to understand how bad the code I am given is, but not enough to make it better.
    2
  20. I think I just came in my pants.
    2
  21. Yeah, I don't need a highly sophisticated stack of transformers and reinforcement learning agents to tell me "drugs bad, mkay".
    2
  22. Go on a vacation and to a festival and get drunk with your friends from home.
    1
  23. I should check out APL, I really like how you can sometimes simply write mathematical expressions in python and julia.
    1
  24. Optimally your institution (university, employer or whatever) should provide you with sufficient compute.
    1
  25. You simply do neural networks and the backprop is matrix multiplications which can be parallelized by default.
    1
  26. Lol, it's more likely for the EU to leave civilization. How about instead people put some effort into coming up with an alternative?
    1
  27.  @ChaoticNeutralMatt  Knowing about yourself IS meta awareness. Maybe you guys are talking about consciousness, which is imo orthogonal to self-awareness.
    1
  28. Nvidia has got us all by the balls. It's time we find some alternative for CUDA.
    1
  29. Should Adapter Factory Controller be in Adapter Factory Factory? Or should it be in Factory Adapter Controller? 🤔
    1
  30. What kind of weird ass boxing gym is that? You get cardio from hitting the bag and sparring and I don't see why push ups and crunches would even be particularly good for boxers at all.
    1
  31. The answer is hashmaps. Unless you're in ML, then the answer is overfitting.
    1
  32. The singularity has arrived.
    1
  33.  @miguelguthridge  The "drugs are bad, mkay" message you get when asking ChatGPT about anything drug related is 100% analogous to a smart person refusing to tell you what he knows, because he doesn't want you to know. It's 100% the same thing. And it doesn't matter whether "feelings" are involved. The fact is that there is a smart answer and you're not getting it. You're getting a generic, dumb response instead. THAT's what's dystopian here. Not that ChatGPT can be wrong and can be arbitrarily wrong for reasons you can't clearly understand. That's true for everything sufficiently complex.
    1
  34.  @kcnl2522  Different CS programs have very different amounts of math. And it depends: If you're not doing anything in cryptography, theoretical computer science, computer graphics, machine learning or scientific computing, maybe you don't need much math.
    1
  35. Yeah I noticed the problem and immediately thought of the WindowResizeLoaderFactoryAdapterFactory class, which dynamically generates WidowResizeLoaderFactoryAdapters.
    1
  36. There is nothing "morally and ethically" wrong with making meth at all. The dystopian part of ChatGPT is that it is so powerful, but pretends to be stupid when it senses that you're doing something illegal or unpopular.
    1
  37. Sure, you can become kind of ok at something easy that you're ill suited for by sinking a ton of time into it, but why? The rational way of skill acquisition and what most people do naturally is to put more time into things, they're naturally good at. Einstein was naturally good at maths and so he taught himself calculus in middle school. He didn't just randomly decide on maths. Same thing goes for sports and programming. When you notice that you're doing better than other untrained people that motivates you to put more time in.
    1
  38.  @emanuel1276  You can use Vim inside of an IDE.
    1
  39.  @umbrella3096  Problem with pycharm is that it's proprietary, so you're only getting the full version if you're a student or you pay for it. And if you ever use another language, you have to re-learn the ide or go for another jetbrains product. VSCode is open source, free and has extensions for all kinds of languages.
    1
  40. Aye, the variable names are half of the program.
    1
  41.  @alilotfi3316  I don't see why anyone would hire you without a formal education. Firstly, I suppose it'll be very hard to get ok at babysitting Deep Neural Networks without peers and professors. Secondly, without the formal education many approaches simply won't come to your mind. I've never used it, but I still remember that various kinds of density clustering exist, for example. And there's not so well known libraries for bugfixing neural network training that probably won't be presented in web courses either.
    1
  42. Seriously though. It's kind of a problem that Nvidia has got us all by the balls. If you're in ML, Nvidia has it's hand firmly grasping your balls with no reasonable alternative.
    1
  43. If you think you suck, you probably do. Stop pretending to know what you're doing and come out as a fraud.
    1
  44. Learning to avoid things is painful, learning to do things is pleasurable. Those feelings coincide with traces of positive and negative feedback.
    1
  45. Not really. All of the above are straightforward consequences of trying to solve certain common real life problems.
    1
  46. Wait a minute... women should be part of being a brogrammer?
    1
  47. Numpy already uses C++ - why reinvent the wheel?
    1
  48. Do your important business via onion routing or sth similar and monero, bought anonymously with cash.
    1
  49. That's entirely accurate. In the same breath in which he mentioned the core count he said that the GPU cores don't function like CPU cores.
    1
  50. That's not how quantum bogo sort works. In quantum bogo sort you suffle the array randomly according to the outcome of sone quantum experiment. The computer automatically checks if the array is sorted and if it is not, a powerful bomb is detonated, killing you. Which means that you experience the array being sorted instantly. However everyone around you just sees you blowing up (in all but the worlds in which it was sorted).
    1