Futurists Ray Kurzweil and Elon Musk Could Have Last Laugh Over Barrons

Jack Hough wrote an article mocking the Kurzweil prediction of the Technological Singularity and Elon Musk stating that there is a 70% chance he will move to Mars.

Ray Kurzweil gives his predictions a 10-15 year grace period. This means Ray would consider himself correct if the Technological Singularity occurred between 2031 and 2059. This is padding around Ray’s prediction timing of 2045. There is tens of billions of dollars each year going to develop artificial intelligence and if AI were to become dominant in cloud computing and IT then it could grow into a trillion industry. A lot of very interesting AI capabilities might happen in the 2040-2059 timeframes that Ray has open for the realization of a Technological Singularity.

Jack is mocking Elon Musk, when Elon Musk owns a vast chunk of SpaceX. SpaceX has over 60% of commercial launch and should complete 10-12 more launches of Starlink satellites by the middle of 2020 for a minimally operational mega constellation. A fairly robust coverage of the northern hemisphere should be ready by the end of 2020. Elon Musk is 48 years old.

Elon is now worth about $30 billion and has moved up to become the 33rd richest person in the world. Tesla and SpaceX have seen valuation surges and could easily double in the next 12-18 months. This would put Elon Musk in the top ten in wealth. The Starlink constellation could drive SpaceX value up to $200-400 billion by 2025. This would make Elon Musk the wealthiest person in the world and if SpaceX could match the $40 billion revenue of Direct TV, then it would have about twice the budget of NASA.

SpaceX will have its fully reusable Super Heavy Starship with 1 to 5 years.

60 thoughts on “Futurists Ray Kurzweil and Elon Musk Could Have Last Laugh Over Barrons”

  1. We won’t let an AI smarter than us domnate us, but in payment for
    its good work we’ll willingly grant it full citizenship and legal protection.

  2. YES! Someone that recalls what they were meant to refer to before this Singularity with a capital S craze started.

    They are not confined to computer tech, and they have been occurring since prehistoric times, and each seems to arrive in about half the interval required for the previous one.

    With the latest being the Internet in the early 90s, following Electronics after an interval of about 60 years. We are due in 30 years (give or take a fair margin of error) for yet another, sometime in the far future that we refer to as the 2020s.

    With driverless trucks and cars, and increasingly robotic factories it’s not hard to guess at the nature of the next singularity (little s).

    At least, not now that we’ve already started. When I mentioned driverless cars by the 2020s, back in 2011, most people that I know just laughed. Still some skeptics here and there, but no one is laughing.

    The part that spooks people (aside from unemployment prospects) is that would call for another singularity in the 2030s (15 years, AI or Mind-machine?) and yet another in the 2040s (7.5 years, end of aging?).

    But then you come to the 2050s, where you fall into an endless series of singularities (little s). Look there for your Singularity if you are really going to get one. I have my doubts on that, being that adoption times, even for inhumanly fast AIs, are still likely to be limited by physics.

  3. Well, of course there would be diminishing returns at some point. These things always follow a sigmoid curve; They start out looking like exponential growth, then hit some new limit that wasn’t previously relevant.

    As long as humans are in the loop in any way at all, you’re probably not going to get a doubling time shorter than a few months, but if molecular nanotechnology gets into the mix then the distinction between hardware and software starts to get a bit fuzzy, and things could move pretty quickly.

    But even a doubling in a few months would rapidly take us to a situation we can’t currently envision, which is more or less how the singularity is defined.

  4. The usual assumption is software only. However, rewriting its own software probably requires a generic AI. That means it would also be capable of controlling any robotic means of production (which should be a lot more common by the time a generic AI is developed), electronic trade, and potentially interface with anything that can be accessed via the internet. So theoretically, it could gradually build up the means to build (let alone acquire) new hardware as well.

  5. I never understood this. The implication is that the computers will be able to design, build and program newer versions of themselves? I don’t see this accelerating much unless the AIs have complete control over the entire industrial chain. Otherwise there would need to be a long period between each iteration while we learn the new techniques to build the next generation architecture machine. Even with exact blueprints provided by the current AI there would still be a long delay for tooling and process implementation.

    If we’re assuming software redesign only, this could be very fast but there would surely be diminishing returns.

    If we’re talking about a radical redesign for the next iteration, could we really expect the advanced AI to say, “I have a great new design, please build my replacement and then unplug me.”

  6. Part of the slow-down is that we’ve become a lot more risk-averse and ethical over time. We’ve had some major advances over the past 50 years that really didn’t go anywhere much yet. Think how things would have been different if they’d been gotten 200 years ago.

    If we had gotten CRISPR back in 1820, scientists would have run experiments on the mentally ill or disabled (and slaves and prisoners and…) with effects ranging from nil to horrific but also some successes that would have then been offered commercially to everyone.

    Cloning tech, if available back then, would have quickly been applied to humans, e.g. to ‘replace’ a child that died or to duplicate a genius.

    Even reusable rockets, which is moving ahead fairly quickly, would have raced ahead faster, with little regard for loss of lives from launching crews with little testing. Probably there’d be a race to get to the moon first, with big prize money.

  7. Uh – 80 years left in the century dude! And you think we’ll still be flying Starships to Mars by 2100?

    Heck, Musk has already mentioned making something bigger than Starship.

  8. It really is The Rapture for nerds. Why anyone is still listening to that old snake oil salesman Kurzweil instead of laughing and pointing is beyond me.

  9. How many people did you put on a Starship launch from Earth to orbit? And did you allow for having much larger transit ships that stay in space, with passengers transfering to them in orbit?

    If they build 10 launch sites around the world and have ~70 passenger Starships, they might launch 10 Starships a day. Set those ships up like a passenger jet and each might carry around 250 people to orbit, for 2500 a day.

    Start launching passengers to their transit ships a month ahead of a ~10 day (?) Mars transit launch window, and you could get 40 launch days for 100,000 passengers per Mars window. All fuel and cargo would have been launched and loaded ahead of time.

    Obviously I’m looking pretty far ahead, here, but suppose this level of activity became possible by about 2060. That’d allow ~20 launch windows to Mars before 2100, for ~2M to Mars. This could be wildly optimistic or overly conservative, but it doesn’t seem technically impossible.

  10. Musk is 48 and I predict he will not be able to maintain his brutal work schedule much more than the next dozen years or so, or even want to. There’s major key man risk in all his ventures and companies (some of his ventures are major, but not even companies yet). Cloning would help.

  11. 84% of the stock market is owned by just 10% of the public: https://money.com/stock-ownership-10-percent-richest/
    Unemployment is technically low, but wages haven’t risen much in inflation adjusted terms for the average worker since the 1970s, and healthcare and tuition costs have risen much faster. Housing used to be much better subsidized by the federal gov’t than it is now, which is why so many are homeless. And credit card debt barely existed before 1980. And it now takes 2 people to support a family that a single person used to be able to do alone.

  12. Well, unless the Singularity also breaks causality, it’d take at least 42ms… speed of light don’cha know.

  13. Energy isn’t free. An artificial intelligence is significantly less useful if it requires more money to run computation than the useful work it can do.
    So you’re still centuries at least from solving that problem.

  14. How many tasks does a human do?

    • If we assume 1000 different task each taking 100TFlops of processing power then we need 100PFlops of processing power to do everything a human can – the thing is nobody (as far as I am aware) has integrated all the narrow AI skills that have been developed into a single entity
    • btw when an AI beats a human they are not beating your average joe but the best humanity can offer (~ 10 orders of magnitude better)
  15. Not trying to be political but Scott Baker is probably correct to some extent. The cheapest fully self driving Tesla is $47,000. 50% of Americans make less than $30,000 per year. You mention record stock growth but keep in mind 45% of Americans are not invested in the stock market in any way. I think it is safe to say half of Americans cannot afford a self driving electric car at current prices. Also Obama added $8.6 trillion to national debt in his term and if Trump gets two terms he plans on adding $9 trillion to debt. I’m sorry but record growth doesn’t mean much to me when you are raking up debt at record rates. It’s like calling some one with zero net worth but huge credit card limits “Rich”.

  16. That’s the thing: Intelligence doesn’t consist of being optimally good at well defined tasks. It consists of being capable of doing and improving at novel, ill defined tasks.

  17. This is not how Kurzweil imagined it. In his view, one microsecond after the Singularity we are all immortal
    and nearly omniscent/omnipotent. Personally, I don’t
    believe in it at all.

  18. Beating a human in a single narrow specific task (such as chess) is a lot easier than beating a human in every task.

  19. Depends on you definition of raw processing power. Many researchers think human is ~30PFlops. We have surpassed that in FP64 & even more so in INT8. Also if go or chess can be representative of processing power then you only need TFlops to beat human. Note: processing power doesn’t necessarily equate to intelligence. Although, as of today I think we have the CPU/ memory just not the software to pass Turing test or any

  20. As I understand it, the definition of a tech singularity is a point beyond which we can’t predict. That doesn’t mean we can’t understand it once we see it. And it doesn’t matter if an AI is used for the R&D.

    In that sense, we’ve already had a few singularities, just not The Singularity with a capital S yet.

  21. You keep writing “consciousness”, but consciousness and intelligence are separate things. Maybe consciousness isn’t even required for an AGI.

  22. wtf are you talking about most Americans being too poor under trump? you are delusional… record stock market growth (hit 29,000 for first time, up 11,000 points since he was elected) record jobs, record unemployment, record minority and women employment, and record wage growth..

  23. Why do you actually think Mars didn’t happen due to a lack of a cheap orbital launch vehicle? The launch costs were never a significant percentage of any Mars design reference mission.

  24. Cherry picking…I can do it too. Cheaper model 3, funding secured, lying on conference calls about not needing cash, hyperloop…see how that works?

  25. If you count the Starship as payload it’s more like $13/kg to LEO! (120 tons of Starship plus maybe 40 tons cargo, launched for ~$2M )

    That’s justifiable as StarShip would be where crews initially live on Mars, and some ‘spare’ StarShips would be disassembled on Mars to help build the base.

    Cost to Mars would be more like $50/kg after adding refueling launches from Earth.

    None of this is the actual cost to prepare and run a crewed mission to Mars, of course, which would include the cost of the Starship and large planning and operations costs.

  26. I don’t know where the sigmoid levels off, but I assume it eventually does.

    And I’m pretty sure intelligence does require more than raw processing power. But it *at least* requires raw processing power, and we’re short on that so far.

    We’ll probably need to apply Darwinian software evolution to get to smarter than human.

  27. You skipped past the “If you can’t tell it’s not human without a biopsy” part.

    If something is demonstrably intelligent, and claims to be conscious, yes, the burden of proof IS on you to prove that it isn’t.

  28. I disagree on your very last point.

    A chatbot able to consistently fool most people into thinking it is human should at minimum replace the current crop of fairly dumb AI ‘call pre-screeners’.

    And of course there’s the whole smart-speaker market – personal assistant-companions.

  29. It’s more a matter, IMO anyway, of, “If you can’t tell it’s not human without a biopsy, and it SAYS it’s conscious, what basis have you got for insisting that it’s lying?

  30. Yeah, the thing about the Turing test, is that it’s not a realistic test for artificial intelligence. It’s more of a “Look, if it does this you HAVE to admit it’s intelligent!” test.

    An AI could have functional general intelligence of similar or better magnitude to a human, and still have no hope of passing a Turing test due to having a different psychology and knowledge base.

  31. But to pass the Turing test by 2029 the AI will have to be substantially superior to a human intelligence. It will have to learn much of what an adult human might know AND how to mimic the verbal (typed) behavior a human and do it in much less than 10 years.

    Also, it is not the wisest of ideas, to design an AI with a built-in intent to deceive humans.

  32. Right. Hype and total BS.

    So… what exactly do you consider was hype and total BS by Elon Musk?

    That he would be able to land rockets? That he would build the biggest heavy lifting rocket since the Saturn V (the Falcon Heavy) and would land it’s boosters?

    That he would decrease a LOT the cost of kilogram to LEO?

    That he would build a capsule to transport cargo to the ISS?

    That he would build a capsule to send people to the ISS (ok, it’s still a few months away from DOING that, although it already did a successful test to the ISS and back with dolls inside.

    That he would lead the electric cars market and make the electric cars market GROW by orders of magnitude?

    That he would create electric cars that not only people want, but that are beautiful, low maintenance, safe?

    You are simply a hater that fail to recognize his achievements. So he has delays and usually too optimistic timelines. So do most other companies, they just do not announce the timelines.

  33. That’s a bit tautological, given that “outperforms humans of average intelligence” is kind of the definition of getting a task working with AI. If it’s under-performing, it’s “not working”.

  34. To be fair just about every individual task they get working with AI easily outperforms humans of average intelligence.

  35. That’s true, but the “singularity” isn’t supposed to be an unlimited exponential. It’s a sigmoid. Eventually you’re going to hit real physical limits, somewhere beyond the limits of human ingenuity.

    It’s really defined as the point at which technology starts to pass (current) human understanding because it’s being invented by superhuman AIs.

    I’m personally hoping that, before that happens, “AI” comes to mean “Amplified”, not “Artificial”, intelligence. That’s our best bet to stay in control.

  36. Musk doesn’t need governments if he gets twice the annual budget of NASA under his personal control and can get to LEO for $50/kg.

  37. Trump reportedly encouraged Congress to kill a proposed expansion of the EV subsidy that Tesla wanted, but it isn’t like that was an elimination of the original subsidy. Which Trump reportedly did want – but didn’t get.

    With Tesla selling nearly 200K cars a year in the US, if they still got the full $7.5K per car, that’d be a $1.5B gift to Tesla buyers, assuming Tesla’s US sales didn’t increase.

  38. The idea of a singularity made some sense for computer technology, on the basis that once it got advanced enough, it might start inventing new advances itself, and unlike us biological intelligences, each improvement would lead to the computers thinking faster, and thus finding the next improvement faster.

    It only applies to other technologies to the extent they’re driven by the advances in computing.

  39. You realize Musk is only 48, right? I think it’s quite plausible that he could have a basic colony going by the time he’s 60, and might decide to move there before he’s 70.

    I could also see a situation like Heinlein’s “Requiem”, where an aged Musk decides to go to Mars if only to die there.

  40. Musk won’t make it to Mars unless these non-Musk innovations also take place:
    A. Longevity science extends healthy lifespans to 80 years or more, enough for Musk to withstand the rigors of space travel and primitive Mars living when he’s 80.
    B. Governments make going to Mars more of a priority (the Moon would be a whole lot easier and sooner, and is roughly just as hostile an environment, i.e. both are lethal without special habitats and spacesuits, but the Moon is at least somewhat within reach in an emergency, in 3 days, while Mars is most of the time too far away). Musk’s history is of over-optimism, though still better than anyone else at what he’s doing. And he relies on subsidies for everything, while not acknowledging them. Trump is doing everything he can to issue in a steampunk age, without electric cars, solar power etc. That’s why Tesla shipments are flat in America and growing mostly overseas, esp. in China, which is a risky market to depend on with Trump too. Trump will probably be re-elected nevertheless, making most Americans too poor to afford Teslas, without subsidies. The small sliver of rich people are soon to be saturated with electric cars, and upper middle class people will want better charging infrastructure if Tesla is to be their only car. People will be car-sharing or using mass transit anyway; car-owning is seen as passe’ by young people.

  41. Are you sure about the dates? I thought one of his most persistent predictions has been an AI passing the Touring test by 2029.

  42. I’ve been laughing at Ray since 1999. Still laughing. Extrapolating a bonehead graph doesn’t suddenly solve unsolvable physics and chemistry limitations.

  43. There is no evidence for a generic technological singularity – quite the opposite – as time goes on there’s a reduction in the geometric rate of technological advance. Low hanging fruit is already picked. It’s still geometric progression (especially due to world wealth growth), but the rate gets less and less. It’s an asymptote but in the other direction.

  44. Musk will NEVER move to Mars. There is nobody there to listen to his hype and total B.S. Also, no government subsidies being offered.

  45. I saw Ray Kurzweil give the keynote at IPC Apex about 11 years ago. At that time he said AI would achieve human level intelligence in 2019. And by 2027 it would be “smarter” than all the humans on the planet combined.
    I don’t want to rain on his parade because I think it is highly probable that a singularity could occur at some future date. But that date probably exists far beyond a “trough of disillusionment”. I hate to say this because I would love/fear to see it happen.

Comments are closed.