Comments by "miraculixx" (@miraculixxs) on "Lex Clips" channel.

  1. Musk's Problem Solving Algorithm is 1. Make requirements less dumb, i.e, reduce to the "one" purpose 2. Delete everything in the system that is not absolutely needed 3. Re-add what is still needed. If you don't have anything to re-add you did not delete enough. 4. Speed it up, make more efficient 5. Automate Often smart engineers start by automating. There is no point in automating something that should not be there. Start by deleting.
    352
  2. Viruses are effectively code. By definition code cannot be intelligent. However code can cause emergent behavior that looks intelligent. I think that's what Lex means.
    69
  3. Good he mentioned the fact that even the gaming models have a human imposed objective function. In other words these models, like all the others, are just mathematical optimizers, and all the ingenuity and intelligence it takes to build them is entirely human. It's literally machine(d) learning, no artificial intelligence whatsoever.
    41
  4. "It's like being surounded by VR porn", Lex: "ohgotitcool" 🤷‍♂️
    38
  5. "I have heard about Ex Machina but I have never seen it", says the guy who wants to invent AGI 🤯
    23
  6. There is some truth to that I guess. Though probably the real reason is some misguided idea along the lines of "real programmers don't use debuggers" which of course is stupid.
    19
  7. I really don't get what people are obsessing about with JavaScript. It's a great language and it gets stuff done.
    18
  8. It's just not true that alphazero et al were not programmed to the rules of the games they play. Read the papers AI/machine learning is not magic. It's clever use of math and computational power. That's all. The intelligence is entirely human.
    17
  9. Dude you alright?
    13
  10. I can totally see why people give him billions. I wouldn't though.
    13
  11. He thinks too statically about models - these are software models, not hardware. What gets sent to cars are a combination of software + data (eg weights + hyper params) and then loaded unto the car's hardware. It is not necessarily the very same model for each car, eg models could be seggregated by geography, region, city. I don't have specific insight, just noting that's possible.
    13
  12. I am not convinced that computer science is going to go away. But I am sure it needs to be redesigned and get it of its brain-dead approach to learning sorting algorithms. What people still need is ability to think, structure, manage complexity, etc. The idea that everyone will be able to think computationally is nice but hardly realistic.
    8
  13.  @matthewvandermeer4934  it took decades of research to get the plate tectonics theory sort of evidence-based accepted. I doubt we know nearly enough about all the other planets and their moons to reach this conclusion.
    8
  14. Funny how Lex has gone full circle from having a pod cast to teach students to interviewing students.
    8
  15. "intelligent" is not a good way to describe a math formula. AI models are just that - math formulae. The intelligence in them, if any, is entirely human, namenly by the developers. All else is just marketing, and not very good one at that.
    8
  16. "Earth is unique in that it has plate tectonics" I'm not sure that's gonna age well.
    7
  17. Lex, thinking that all it takes to end the war is for Zelensky to agree to negotiate is naive on first and second order thinking. Get your facts straight.
    7
  18. Sam is such a pleasant person to listen to, I could go on all day. 😮
    7
  19. Scala? Hm. Who uses it? And I don't mean the great folks who created and maintain the language and its ecosystem.
    6
  20. What I enjoy most is young people telling us we absolutely must go vegan bc reasons and then proud themselves of wearing the newest sh* they ordered on Etsy or something and had it delivered over night straight from some Southeast Asian sweatshop. 😂
    6
  21.  @Ian-eb2io  "no way" is impossible
    6
  22. Why would something smarter than us kill humanity? That doesn't sound intelligent, it sounds recklessly dumb. I'm not afraid of AI, but I am concerned of the stupidity of human operators.
    6
  23. "Formal verification" for everything really gave it away. Tell me you are clueless without telling me.
    5
  24.  @helgeh  they do important(tm) work 😉
    5
  25. Mainframes are not slow!
    5
  26.  @HeavyDist  thanks for the feedback. I was not trying to prove anything.
    5
  27. In a nutshell: Language models were introduced some ~15 years ago, i.e. models that can generate text. While they generated text, these were not very good or useful. Several smart people tried different approaches (RNN, WaveNet, etc. finally Attention/Transformers), and ultimately found a model that works really good, but on a small data base. Google, OpenAI, and some others, were in somewhat like a research competition of getting better and better models, using more and more data. Then OpenAI was bold enough to use all the data they could get their hands on. And that gave us ChatGPT.
    5
  28. "What is the best movie?" - "The Godfather. Btw I just did a movie with the Godfather's creator. I met him. I was with him on set. I made this movie."
    4
  29. Intelligence goes far beyond text generation. Yes these models can simulate(!) human-like reasoning but it does not actually think.
    4
  30. ​ @raphaelamorim same. Also programmed soc with 64kb to do pattern detection just using plain c code.
    4
  31. Chomsky would accuse the US of expanding its might and causing downfall of the world, Wolfram would say there is perhaps a fundamental law that makes the language of Chomsky sound the way it does. 😅
    4
  32. What?
    4
  33. Thanks for that! Observing the same. Also driving a car does not take that much consciousness once you get past the learning phase. In other words self driving cars are not a sensible example of AGI-in the making. Rather it is a sophisticated machine with a very specific purpose, making many routine observations and taking conclusions to act. Impressive, yes, not intelligent and certainly not concsious.
    4
  34. He doesn't come across like very thoughtful or otherwise qualified to deal with society-wide technology.
    4
  35. If somebody tells me about blockchain this and that I replace "blockchain" with "database". It usually is just BS after that. Same thing with NFT. Replace NFT with "payment". Keep it sane.
    4
  36. Yes. Humans have been living in space for at least 300'000 years. Duh The more interesting question is, why did it take us 299'950 years to build a capability to visit other stellar bodies? Or did it?
    4
  37. Combine many important sounding words and you'll have yourself a talk with Lex 😮
    4
  38. With all due respect I don't think this guy knows much about the technical details of GPT and other AI models. There is literally nothing in these models that could even potentially develop a deceptive conscioussnes, nor any other kind of conscioussnes. It's just maths.
    3
  39. Perhaps we need just one additional law: "Any AI system, for all intent and purpose, is treated as and subject to the very same rules, laws and societal pressures as human beings."
    3
  40. It's the Dolphins they are interested in. Obvious!
    3
  41.  @shark23-17  come again?
    3
  42. We need a minimum age for pod cast participants.
    3
  43. I don't see how these youngsters can appreciate the complexities of real world IT systems. Naturally they can't. They use terms like "engineering department" but in reality they don't really understand neither of these two words. Engineering in the real world is exactly not about code, but about humans collaborating to make a systems work. The way they talk about this is very revealing, but also very common for young folks.
    3
  44. Doesn't seem to like Intel that much :)
    3
  45.  @Pat315  actually you do. If you don't, why a program works or rather why it does not will stay hazy forever
    3
  46. If he thinks he can get Putin to back down why does he not get on the phone and call the guy?
    3
  47. Yeah except in Germany, if the recent pandemic is any guide. They would immediately have about 30% of the population arguing that the missiles are not real, and even if they were that "getting hit by one is good for your immune system" as well as "dying is something we have to live with" 😢
    3
  48. There are also very few systems that need that scalability, duh
    3
  49. Vscode sucks compared to PyCharm. It keeps amazing me how people jump to the most recent coolaid instead of using the tool they like
    3
  50. The first technology was likely not fire, but sticks or stone. Mastery of fire as a tool came much later. She says stuff very beautifully and with authority but that doesn't make it right. 😮
    3