Comments by "Michael Nurse" (@michaelnurse9089) on "Wes Roth" channel.

  1. Can hold fast conversations. Underlying LLM is like GPT2 level.
    92
  2. Only in California. Everywhere else sanity prevails.
    35
  3. AI (and later robotics) is going to kick 90% of professionals in the nuts. Try to think who is safe and you get a very short list.
    18
  4. This is why I invested the bulk of my savings in Tesla. It is a volatile stock with a volatile CEO but nobody is better placed to profit off AI than Tesla. The 5-10 year outlook on Robotaxi and Optimus is incredible.
    18
  5. Altman purged everyone of value.
    18
  6. You get paid when you vote for Sam in 2028
    14
  7. The modern music industry is anti-musician. The problem is that this will kill both the music industry and the musicians. @flickwtchr 
    13
  8. Internet became dead a long time ago. I am not actually typing this, my commenting bot just auto-responds to comments like yours with bits of text like this.
    13
  9. You called it ChadGpt. That name might be better.
    9
  10. Silicon Valley bro accent.
    9
  11. Its the national anthem of the largest nation on earth, the internet.
    9
  12. He said something that was very, very, very slightly right leaning. Like people should work hard.
    8
  13. Sure. But, but the Apple R1 will sell 2 billion units.
    8
  14. Everyone in tech is on the spectrum. It is just a question of how far 'down' they are.
    7
  15. Answer: Higher quality data. (Overfitting is something else - it is where the model learns a bunch of specific cases but can't generalize well)
    7
  16. Same thing happened with the first two industrial revolutions. The story of the generation involved is told by Charles Dickens. Later generations were fine.
    7
  17.  @blinkers88  Right principle. Bad example.
    7
  18. Competitive programming is well suited to LLMs. Creating agent programmers that do the 'whole job' of a programmer will be very difficult and very expensive.
    6
  19. ChatGPT can rap like Eminem: Verse 1: Yo, let's dive deep into the neural net lore, Talk 'bout a genius, Schmidhuber, and more. In the AI world, he's a hidden king, Crafting deep learning, making machines sing. 1997, yeah, he changed the game, With LSTM networks, ain't nothin' the same. Long Short-Term Memory, that's the key, Unlocking secrets of sequential data, see? Chorus: Schmidhuber, Schmidhuber, deep learning's unsung hero, Crafting layers, neural flows, from ground zero. In the shadows of Hinton, LeCun, and Bengio, He's the silent genius, the real deep learning maestro. Verse 2: From perceptrons to layers that go deep, He navigated paths where angels fear to creep. In the AI winter, when hope was thin, Schmidhuber kept dreaming, never gave in. Recurrent nets, man, they were his thing, Making them remember, making them sing. A brain-like machine, that was his vision, Cutting through complexity with surgical precision. Chorus: Schmidhuber, Schmidhuber, deep learning's unsung hero, In the lab, grinding hard, far from zero. Behind the scenes, making AI dreams glow, The real genius in the deep learning flow. Bridge: He's not just a name in the academic papers, He's the architect, the code, the hidden shaper. In the world of AI, he's the quiet storm, In the deep learning saga, he's the unconventional norm. Chorus: Schmidhuber, Schmidhuber, in the AI hall of fame, His contributions, his genius, we gotta acclaim. In the tale of deep learning, let his name echo, Schmidhuber, the mastermind, the real tech virtuoso. Outro: So here's to Jürgen, the legend, the sage, Writing the most crucial page. In the epic of AI, his story's a must, Schmidhuber, the genius, in deep learning we trust.
    6
  20. Reduction in Hallucinations addresses a massive problem with LLMs. Wes just dismisses that as progress to get some clicks.
    6
  21. Tim Cook was drinking his Vegan Chai Latte and it came out of his nose...heard from a friend.
    5
  22. I don't think that is the plan. The plan is instead of having 3 senior and 3 junior PHDs you run with 2 Senior and one of these agents (who can work 10x the speed). Think 2030 and not 'July'.
    5
  23. I can see Sam suing South Park shortly thereafter.
    5
  24. I also like how Sam Altman is predicting future profits of one trillion dollars. The delusions of grandeur are simply staggering.
    5
  25. Ultimately, big tech have a strong moat, because after training on the internet data sources are proprietary - scientific papers, books, internal business documents, medical records, emails - you have to pay for the data. Open source providers may struggle to get that data as they lack the resources of the big players. Also, training costs keep going up - they are not YET a moat but five years from now it may cost $10bn to train the largestt model - and that cost has to be recouped from customers or elsewhere.
    5
  26. probed up the exit hole
    4
  27. Meh...show the proof that Elon is anything other the World's best engineer. S=fastest car. Y=best selling car. CT=most preordered car, and I never even said the words rocket.
    4
  28. They are not saying that. They are saying they invented a tractor for programming. When the tractor for farming came 99% of farmers were unemployed and left for the cities - the subject of every Charles Dickens novel. 1% of people are still farmers today - they survived just fine - just in lower numbers.
    4
  29.  @OriginalRaveParty  This is important. There will be a small group of 'heavy duty' programmers that will be unaffected much by AI. The group of 'do some programming' will 10x with LLMs as they improve. The smaller groups jobs are safe as LLMs cannot threaten them. The larger group's jobs are also safe for now but those who cannot usse LLMs to program and automate are threatened.
    4
  30. There will be millions of AI jobs. But with 5 billion needing employment there should be a job for every 1000th person.
    4
  31. That what the delay is for. So they can move the drone nearer your location.
    4
  32. Sometimes when you stare into the dread it stares back at you
    3
  33. Ozone is toxic if inhaled. Sam Altman is toxic even if not inhaled. I get it now.
    3
  34. Also start paying for exclusive purpose designed data. Would work well for medical advice/diagnosis and for programming.
    3
  35. Sora will be sold to Hollywood so never coming to you. Voice probably similar. $20 subscriptions can only go so far.
    3
  36. They are going to need a lot of electricity, so all Canadians will have to move to Mexico to make space for that. ¿hablas español?
    3
  37. The device has the wrong vision. It needs to control all your other devices using natural language. "Send a message to John asking if we are still on" or "Run the reports again for 1st quarter only" or "Check what they kids are up on their phones" or "Put on Netflix on the office tv".
    3
  38. Was thinking this. And Morgan Freeman for male.
    3
  39. That is the question here - how do they get their data fly-wheel going? No doubt they share it - perhaps not directly - but via a training process.
    3
  40.  @DonG-1949  Rhetorical: If a bot speaks some words, and they happen to be true, should they be disregarded out of hand because a bot was involved? The internet is increasingly dead but you will never know who is real. You could be a Sambot. I could be an Elonbot. I could be a Sambot pretending to be an Elonbot to make strawman arguements.
    3
  41. Got me too.
    3
  42. They never said WHICH Thursday. Think about it for 1 second.
    3
  43. Well, a calculator is smarter than any human, at arithmetic - but it has no real will of its own. I wonder if it is a stretch to apply this same logic to AI ? No idea.
    3
  44. "In war aggression wins" - this is schoolboy talk. In WW1 and Ukraine defensive systems had the advantage.
    3
  45. The greatest shortcoming of those who understand the exponential function is to realise there is no such thing, only s-curves.
    3
  46. I suspect you are just an AI troll bot.
    3
  47. Sensitivity is a spectrum, from HSP's all the way across to those threatening to kill puppies on every prompt. Everyone seems to advocate for matching their own level of sensitivity or their political bent. It just needs a red/blue toggle switch and a sensitivity slider.
    3
  48. You would have to retrain its neural net each time - and currently that costs many millions and take a couple months.
    3
  49. Their AI is open - open in the sense of 'open for business' at $20 a month. No need to rename anything.
    2
  50.  @davidantill6949  I am imagining Snatch right now.
    2