Comments by "Michael Nurse" (@michaelnurse9089) on "David Shapiro" channel.

  1. Dave "AGI cooming in Sep" AGI enters left stage Dave "I am not impressed"
    106
  2. "I am sorry Dave, I can't do that." - ChatGPT6
    50
  3. 1. Dave gives up doom. 2. Dave forgets he taught them them how to create artificial data. 3. "Open the Bay Doors, Hal" 4. "I am sorry Dave, I can't do that"
    43
  4. Manhattan Project 2 is currently underway.
    21
  5.  @woptropical  I saw this coming. Dave was basically dragging himself through the o1 release with zero enthusiasnm. That was my impression.
    20
  6. Those who speed it up to control it, however, are 100% trustworthy. Sam Altman, I am looking at you.
    15
  7. I don't see these as issues: the real issues are: 1. The frontline models have too low accuracy. Hallucinations are everywhere. 2. Costs to 'create a whole business with AI' are high. Time spent/people paid + token prices = a big investment.
    15
  8. Prompt: Views are down on my channel. Do mw a script for a video that will bring out all the viewers. Dave would never do this though, he is a high morality person.
    13
  9. This: Never underestimate how low narcissistic bosses will go when they get new technology in their hands.
    8
  10. In the future, you will be able to work 100 jobs at once, just make sure you buy high quality AI agents and keep all your bosses satisfied with regular AI phone calls.
    6
  11. Depends. In Africa, grassland not eaten up then defecated by cattle quickly turns to desert. Also, World population may decrease - it certainly is among richer countries.
    5
  12. No. Tesla is the fastest production ramper. 2025 - finalize v3 prototype. 2026 - build factory - 2027 - ramp factory, 2028 flood workforce.
    5
  13. Militaries turned out to be very controllable. Uncontrollable militaries usually engage in a coup d'etat. When was the last such thing in a Western/Democratic developed country? (Greece 1967)
    4
  14. They don't announce when something is training or not. People are just guessing. You can partially train, stop, add more data, continue.
    4
  15. When OpenAi re-incorporates as for profit it should rename itself. DeepstateAI might be in line for the suggestion box...
    3
  16. "Models that say they have a context of 32k+ tokens end up forgetting or ignoring instructions after as little as 15 pages of text," This
    3
  17. "The best defense is a good offense" this might apply on the basketball court. In the real World, if the US build an unstoppable bomb the 'enemy' steals the design and builds it a year later.
    2
  18. If I do a water fast or cleanse my anxiety evaporates. But it never lasts. It comes back, perhaps milder. Fear can be a faithful tool (when the danger is real) or a disability (when the danger is not real).
    2
  19. "AI will kill everyone" is ridiculous. That "AI will kill significant amounts of people" is less so. Ukraine desperately needs self piloting drones that can operate while jammed.
    2
  20. OpenAI will release a new model every 18 months. Like Intel used to do.
    2
  21. "We've been accelerating on a parabolic curve for the last 100 years." Depends on the subject matter - physics has made no significant progress in the last 40+ years.
    2
  22. You would think so, but those actually doing this sort of consulting are struggling, or at the very best doing as well as a lawn care company.
    2
  23. Robots are some way away still. All the companies are in prototype stage including Tesla. Expect to start feeling the impact in 10 years if my gut instinct is correct.
    2
  24. I don't think this will be a thing - Tesla has plenty of things that go wrong with FSD but they just add better data and the problem goes away - no need to track down the cause.
    1
  25. Capitalism will cause problems for AI. Communism, even vaguely similar to its 20th century USSR form, when combined with AI, would be the end of everything we know as good.
    1
  26. Inference is limited by single thread speed and that has been flat for more than a decade. I can imagine future LLMs which take 24 hours to get back to you but accuracy is excellent.
    1
  27. Tesla can build a mostly automated car factory in two years (1 year in China) and ramp it up in another year. Because Robots are new it might take double, so six years. As for raw materials you get about 100 humanoid robots for every car. The most importantt thing is getting the software to work well.
    1
  28. Fusion - there are those claiming they will 'bring it online soon' but they are generally searching for investment - so I trust their opinions about the same as a 2nd hand car dealer. Some industry veterans are saying we will still be 30 years away in 30 years time.
    1
  29. Hot war with China. How hot? Thermonuclear hot or just a couple of Amraams fired off the coast of Taiwan?
    1
  30. Larger compute coming online could also be involved.
    1
  31. That's what she said.
    1
  32. Open data + open weights + open reflections + no political censorship = open ai. Best to worst in order. Grok (X), Llama (Meta), everyone else, last place = Google.
    1
  33. Covid was a bioweapon. How did we do?
    1
  34. This is all fine and well. But not all transitions are equal and WE are collectively the authors of that transition. Imagine that post Manhattan Project the US took a different path - to pre-emptively nuke anyone else who could build a bomb. This was a real option at the time, but a more sane 'transition' was adopted, probably for the better.
    1
  35. Axiomatically, you start with 'suffering is bad' - I am however reminded of the Universe 25 experiment where removing suffering from the rats resulted in eventual massive overpopulation that then resulted in social breakdown - rat on rat violence. I think humans are already about 7 billion past this point and acceleration from here will lead to something akin to the movie Idiocracy.
    1
  36. OpenAI is keeping ChatGpt-5 under lock and key until they are certain it will bring good news cycle publicity.
    1
  37. No. Once humans are economically useless we become a liability. Either they give us a food and entertainment allowance or find ways to kill us. The primary customers will be other corporations with their machines. Customers are not really needed when you control all the resources - you just use them directly to build your empire - no need to buy and sell stuff anymore.
    1
  38. Tesla's first bot is targeting $20000. I know you don't like Tesla because of Elon's politics.
    1
  39. If you run for president I am voting for you.
    1
  40. The problem with spinning up a trillion agents is it will cost a trillion dollars and nobody, and I mean NOBODY, has a trillion dollars.
    1
  41. I think Dave might be getting to far over his skis here - shortage of free data could put a stick in the spokes of the whole thing.
    1
  42. As an investor I can tell you 99% of people today are thoroughly incapable of being a decision making investor. I would go so far as to say that 90% of professional decision makers are just herd animals and so are also unsuitable. There has to be a better way than creating a giant casino where every beer drinking Joe can gamble his life's assets.
    1
  43. Picking AI assets managers changes nothing - you will pay them for their service and those fees will eat into returns. Just like picking meat asset managers today is typically a path to underperformance. Far better is the index method - everything goes into an index and everyone automatically owns 30% of the index - 5 year rolling average growth in the index can be drawn out in cash after leaving a buffer behind.
    1
  44. Neither. AI cannot surpass the intelligence in its training data. Instead it will get better and better at driving up corporate efficiency, military efficiency, cyber attack efficiency - which can cause problems that ruin your life all the same.
    1
  45. To mine Moon, Mercury, Mars, Asteroid Belt we would need roughly 1 trillion - so 50%.
    1
  46. Tesla has declared plans to 10x Optimus production each year. Elon said no guarantees but if it goes to plan it looks like this: 2025 - 10k 2026 - 100k 2027 - 1m 2028 - 10m 2029 - 100m 2030 - 1bn At which point every business and every 1st World family will have one.
    1
  47. Cars were curiosities before mass production. Cars displaced horses because Henry ford reduced their cost 10x. This made cars many times cheaper per mile than travel by horse. People never cared about the inconvenience so much as they cared about the cost.
    1
  48. It is a fallacy that if consumer demand drops to zero the economy grinds to a halt. There is still business demand and government demand and exports.
    1
  49. The common heuristic today that consumer demand = economy is first of all a US thing and 2nd of all just an illusion that results from the US consumers spending a lot of money on a lot of junk. The implications of an economy that only produces junk and only consumes junk is well documented in the film Idiocracy.
    1
  50. Tesla is the only AI stock with a clear path to monetization. FSD took 10 years and data from 5 million cars to get to current levels so it has a moat.
    1