Elon Musk Is Using Trillion Dollar Power to Create AGI

73 thoughts on “Elon Musk Is Using Trillion Dollar Power to Create AGI”

1. Baby Yoda? Happened to just see Cleese explain how being British was to be terrified of humiliation, to Wanda. More seriously, long ago I heard a green guy explain that the popliticians would always vote in their re-election interest except in one case: If they thot they were going to get caught humiliating themselves. Mentally ill neurotic power addicts have something to worry about.

2. China banned Winnie the Pooh cartoons (at least for while) because of a couple of people depicting Xi as Pooh.

   That's a snowflake so delicate it must have been grown in zero-g.

   Meanwhile, in the People's Republic of the USA, cartoon frogs are not exactly banned by the federal government, but can get you punished by government institutions such as universities.

3. They don't know, they don't care, and so they don't whine and demand a pay increase.

   I'm not seeing that abandoning this happy state of affairs will be a good move.

4. Are you calling the leaders of major countries "delicate snowflakes"? I agree!

5. But do they *know* they are doing those things?

6. Rabbit Hole from Alice, rat race from 50s, lemmings?, blind alley, dead end. I'm not sold on androids, but robots in more abstract sense, more like lounge chairs? edit: Run Around!

7. Ah! Makes sense now.

8. I have machines that will do the dishes and laundry already. Not new actually.

9. China is a big enough, and aggressive enough, player that they can (at present) enforce their rules over their land area.

   And I suspect that USA and Europe are also, though for much lesser values of how strict their rules are, and what methods they'd use to enforce them.

   Ie. Today you can access all sorts of completely uncontrolled internet in western countries. But it's been pushed out of the family friendly areas. Twitter and Facebook and Amazon are "controlled" because that's where your typical unsophisticated user would accidentally, or "accidentally" encounter something scary and result in politically damaging headlines. But venture into even fairly mild areas like 4chan and you're on your own. Let alone the really wild dark web. Karen from HR will never go there, so she won't be complaining if there is something naughty there.

   BUT (to get back on topic), just because China and Germany and California stops you from offering anything that offends their delicate snowflakes, doesn't mean that Venezuela or Syria or Liberia will be able to have any say in the matter. Nobody gives a damn that your free satellite web is saying nasty things that the Laos government wishes were kept quiet.

10. Yeah it went kinda fast from "oh cool wow, you are the man elon" to "you own most of the space stuff, most of the tech, what do you want to do with it?, Can we stop you?"

    And those kind of observations are made on progressive futurist forum, some conservatives must be peeing their pants every time when they see Musk article and they arent wrong

11. At a guess, Musk is going to be completely pragmatic with Starlink. Every nation that wants to control what their citizens can see/hear/do will have their own Starlink uplinks from their own internet. If Starlink eventually provides orbital hosting, either that won't be used in those nations, or it'll somehow be firewalled.

12. "Elon is using the power and resources of a trillion company to drive the creation of AGI (Artificial General Intelligence). "

    If he were serious, he'd increase Marcus Hutter's underwriting of the Hutter Prize by a factor of at least 100. But he's not serious. He's already caught the court toady disease.

13. There are many such studies. The reason "why" Space is that it is better and we can then restore the Earth to Nature, and repopulate with Neandert(h)al as reparations.

    https://space.nss.org/space-settlement-library/

    edit: https://www.youtube.com/watch?v=tZcTYFrOzos

14. What are the reasons and feelings of people on a decision between space or planetary location, wrt to security and stability (directly, not on a 2nd, 3rd generation located there already)? People are fascinated with space movies unless it would become a real necessity? Sounds similar like getting used to artificial companions like robots on a new habitat. Did psychologists influence engineers/technicians planing new space/planetary locations for improvements?
    Did anyone ever quantify the amount of parameters and qualify their importance for a comparison between space and planetary locations?
    Seems a big question (within comparison of optimized concepts) is related to "why" places are going to be colonized?

15. "materialized in space/micr0g. . .human society without 'matter'" sounds pretty far out! If you are thinking of Space Settlements and such built of matter, O'Neill does exactly that.

16. Indeed. Things are going my way on O'Neill. We seem to be going to the Moon despite Musk's Mars First/Direct/Only mindset. Big NASA teleconference at 11 CST on this!

17. I think Musk is a planet chauvinist, and that it is clear that building stuff in cislunar and towing it out to Mars orbit or a little further, as you like, then perhaps adding heavy stuff like radiation shielding, is FAR easier start than Mars surface, for any given size. Unless, by exclusive postulation, you can only live on a planet. Is the surface of a planet the right place "for a civilization preserving colony" that is "far enough away from Earth that it wouldn't plausibly be taken down by the same extinction event."? If the answer is clear to this, it is very important! Remember, O'Neill Asteroid Space is better than Earth, not even considering lifeboat stuff.

18. I was joking, as was XKCD; Even if you've got the bandwidth to pull it off, such situations come up too fast for a handover of control to a human. People need a few seconds to come up to speed on a situation.

    You *might* be able to have a 'cloud' based system where edge conditions get handed over to a system that can devote much greater computational resources than the individual cars have, and then solutions get incorporated afterwards into the cars' neural nets. With the car driving more conservatively when the connection isn't available.

    Or, similar to the way Google handles traffic slowdowns, if you had enough Tesla cars on the road, the ones that happened to be manually driving when they encountered the condition could serve as a resource for the auto driving ones to know what to do.

19. Musk will have to limit that at first, because he won't want China shooting down all his expensive satellites. But Starlink should enable bypassing the less aggressive censorship that's becoming endemic in the West, because it's sort of informal, and at least in the US they pretend it isn't being driven by the government.

    Especially once Musk offers orbital hosting, which seems an inevitable development for Starlink. That's got to have the NSA sweating, because what good are their taps on the ground based internet switching centers, if end to end encryption traveling overhead becomes popular?

20. It should be all of those things. Successful species and civilizations don't do just one thing, they branch out into all the options.

21. Yes, that could definitely shift the economics in favor of earlier orbital habitats.

    It's not so clear that it would change Musk's plans, because one of the criteria for a civilization preserving colony is that it be far enough away from Earth that it wouldn't plausibly be taken down by the same extinction event.

    Any colonies in the Earth Moon system stand a good chance of going down if something takes Earth down. And the economics of being so close to Earth would discourage genuine self-sufficiency.

    Mars is far enough that the trip makes for a fair quarantine for plagues, shipping costs will encourage self-sufficiency, and a world ending asteroid wouldn't catch them with the shrapnel. It's about ideal for Musk's purposes, if only for them.

    I'm an asteroid colony guy, myself, but Musk's work is going to make that feasible, too.

22. No, but SpaceX is launching more rockets than everybody else combined, just like they have more (partially) self-driving cars than everybody else combined. And they're doing rolling innovation, using this to learn things.

    Building up hours of experience is important, it causes various cases to happen often enough to collect usable data on them.

23. Has anyone ever done a study/energy balance comparing a colony materialized in space/micr0g or on planets within 'our' solar system (and tradeoffs in between, optimized on its purposes for support of resources and abilities/time)?
    What's the value of human society without 'matter' (beyond life support)?

24. Something like the internet,

25. Ha! This just in: "If we are able to discover more Earth Trojans, and if some of them can
    have orbits with lower inclinations, they might become cheaper to reach
    than our Moon," Briceño said. "So they might become ideal bases for an
    advanced exploration of the solar system, or they could even be a source
    of resources."

    https://www.space.com/earth-extra-moon-trojan-asteroid-2020-xl5-discovery

26. As an African, my concern is that we need to give reparations and restoration to the Neandert(h)al. And make up for a lot of other stuff. See Alice Miller "For Your Own Good" for your specific historic example. Hitler's childhood trauma, and his generation's, led there also, perhaps mostly. Janov has the mechanism on display.

27. Actually, if you read the text, he is describing classic bipolar. The rush of ideation. I did read somewhere that the name has become subsumed into "neurodiverse" because of this exact sort of superiority vibe. I, on the other hand, am better than others at being humble.

28. You are taking a similar topic as Isaac Arthur, he saying there is not enuf material in orbit to suffice. I agree with the big picture!

    The habitat mass is almost all radiation shielding either way, is it not? I am saying that the effort to do things, even the small stuff like habitat for the humans, is far beyond wildly harder in a g field, even Moon's. Getting the material from that yet to be discovered set of small NEOs with the Queen Bees will show how efficiently moving mass can be done. In micr0g. From then on, drop the cranes. Dial up the g you want, big, small or micr0. Have some free vacuum and sunlight. Plenty of room. ISMRU. Do it! Then, tow the working stuff out to safety, far before it can be done on Mars. Tow it to Mars.

    And this can start with launched raw materials. Micr0g will allow ISM product for Earth and Space use that is now impossible. And construction of simple things that are complex to launch, needing folding and sturdiness for nothing but the launch. $$$$$$$$$

29. I totally agree. "micr0g will be done by others than Musk" is my point. He will not be doing micr0g. I have said many times I hope his rockets are soooo good that Bezos throws in the towel and buys them. However, micr0g is not an option, it is the future. Forget Mars. Just because Musk likes Mars does not mean it is the word of the gods. See the almost daily NASA TV ISS crew interveiws with locals, etc. They are not excited about going to a place where this micr0g stuff cannot be done. Should the upcoming lunar base be primarily crew in orbit, visiting the Moon, or primarily on the Moon surface, stopping in orbit just to change vehicles and refuel? Where will we experiment upon and process rego?

30. Agreed. For life on space to become possible, we need cheap reusable launchers first and foremost.

    Without that, it's just literal building castles in the air. Human space already has too many PowerPoint dreams that go nowhere.

    Musk is obsessed with Mars. So what?

    Let him build his Mars city and bring whomever wants to go! it's their money, their problem(s) and their lives.

    The rockets, spaceships and infrastructure they build to make that foolish dream happen also will serve any other parallel purposes than life on Mars.

    We are currently devoid of the means to do anything like that happen. Expendable or slightly reusable rockets as they are simply won't cut it. We desperately need something that bring payloads cheaply, in bulk and repeatedly.

31. Prefer this approach in construction and war zones, but not for L4 cars carrying your nanna.

32. Neuralink = Hope.

33. I'm guessing it might be more like a many-to-many rather one-to-many transfer relationship. A robotic society of self aware Telsa cars and bots might just emerge!

34. "Ubermenschen" is a bit of an exaggeration when talking about Aspergers. It's more like a handicap with compensations, than a superpower. Sure, you get a basically unlimited attention span and a greatly improved head for details. But being a member of a social species and lacking an instinctual grasp of social relations sucks.

    The increasing use on the left of "autistic" as an all purpose insult is annoying, but spare me the superman illusions. I'm glad my son didn't get my Aspergers; He might not be effortlessly acing all his math classes, but he's happier than I was.

35. Look, let's try to think about this rationally. WHY are O'Neill colonies, or something like them, the long term destiny of an expanding species? Because planets have limited area, and require about 10^13kg/square meter to achieve Earth gravity. O'Neill colonies require about 8^3kg/square meter the achieve Earth gravity. So the same mass can provide billions of times more living area in the form of O'Neill colonies.

    Planets don't scale.

    But this isn't an immediate concern. It isn't going to be a concern for hundreds of years, maybe never if we don't fix that birth dearth.

    If you're on a planet that already has gravity that's acceptable, the mass necessary to produce a livable habitat is more like 8kg/square meter. Advantage, planet. Until your society won't FIT on a planet.

    Musk isn't trying to create the future of a thousand years from now. He's trying to create a backup plan for Earth in case an extinction event happens, and he wants to do it as cheaply and as fast as possible, because we don't know WHEN that extinction event happens, just that in the long run there will be one. Yellowstone could blow up next week. An asteroid could strike the Earth next year. Musk doesn't have time to be farting around with the best long run solution, he wants the solution he can implement NOW.

    For his purposes, Mars makes sense.

    Just be happy that his purposes require providing all the stuff that's also necessary to build O'Neill's future, too.

36. They are reacting to the air of superiority "ubermenschen" such as we have, clearly being more fit for a tech world. However, many on the same (pc warning) neurodiverse spectrum do have severe problems, so it is more a matter of luck to hit the sweet spot, attention without total absorption. Or bi polar.

37. There are no other rockets, and no other AI exists in the real world. Really?

38. What if you want to develop Space? Only if you *assume* a planet goal will you think going to a planet will help. Maybe to get resources. Also, self sufficient starts with productive, then grows to self sustaining productivity. Mars has no way to get support from Earth by selling anything much. I know, reality shows will be the only way to make money, "in the future". For now, Al Globus ELEO plans will support millions in full radiation exposure, where they can make and sell micr0g stuff, to Earth and others. Long before any such numbers on Mars, always far more than on Mars. Forget planets!

    You may find that covering the balloon, I assume you do not mean floating, with stuff takes more effort than bringing stuff from Bennu, or the Moon's surface, to cislunar. You have to dig, move and carefully place, each in .4g, which adds up to a lot of equipment. Easy in micr0g.

39. The similarity to what we actually have is striking. We actually have a system of replicating ritual and behavior that is involuntary and unconscious. It does not live in our robots, but in our *souls*. It operates with the mechanism of repression, present but only normally developed in chimps. It is a life form similar to bacteria, no real individuals but able to evolve. It evolves by selecting the rituals that lead to more of the self-same ritual over others that do not. This is usu thru the infliction of more childhood and birth trauma than the competing rituals. Have a nice day!

40. With decent bandwidth, all the system needs to be able to do is recognize oddball conditions, not solve them. It can refer them to 'drive centers" located in the third world.

    https://xkcd.com/1897/

41. A good depiction of this future, if we are (un) lucky, is in Alastair Reynold's "Blue Remembered Earth" novel.

    There the future is controlled by several AI/human polities, the biggest ones being the Chinese and the Mechanism, with others like the Merpeople, the rogue quarantined AIs on Mars and a few more across the Solar System.

    The big ones were sustained by AI systems regulating the public behavior, watching people's (and some animal's) every move, and even stopping them through brain implants, which are supposedly only used in extreme necessity (like against any display of violence against humans), practically removing all crime and violence, but it's left clear in the books that in exchange for that, privacy and complete free will are a thing of the past.

    The Mechanism is depicted as able to plan and perform great engineering feats across the Solar System (it has all the decades old self replication and building recipes) by itself, and people can simply choose to go live into these settled places without any inconvenience like having to build it. The world has been in such a sedate state for a few centuries already, cause they aren't risking their necks doing stupid things like going to others stars, despite having the requirements and tech already. The world is finally happy and content… and stagnant.

    Of course, the Mechanism turns out to be not really that benevolent, and it turns out to be serving its own self generated AI agenda.

42. If you want a self-sufficient colony in the near term, on Mars you can live in a balloon. In orbit you must deploy 7-8 tons of processed material per square meter of living space, not 7-8 kilograms.

43. Like with their rockets, Tesla has the advantage of real world feedback.

44. Arthur Janov wrote "The Primal Scream" about the same time O'Neill asked his question. Both are essential for the future. primaltherapy.com

45. Where do people fit into the $Quadrillion economy? Will we be limited to our little bungalows built by robots eating synthetic food from factories? Might be comfortable and tasty, but I wonder if being parasites in a cybernetic dominated economy is where humans will end up. The androids go into space and build huge mysterious edifices…and we humans tag along for the ride…inhabiting the nooks and crannies where we can siphon off the resources going to unfathomable projects created by alien minds.

46. Everything Tesla makes is clunky and 90% complete – or less.

    The cars are plagued with quality issues. The UI software is slow and unresponsive. The auto steer kills people by running into parked objects. The driver assist regularly throws false alarms about dangerous braking with certain designs of brake lights.

    Why does anyone care that they're also going to produce incomplete and unreliable robots?

47. On Criswell specifically, it is the exception that proves the rule. There is sooooo much soooooo special about the Moon that it is an astonishment. Just heard that tidal heating forces of the early close Moon kept life alive. Our type star otherwise starts too cold or gets too hot. Criswell did argue against launching SPS product from the Moon, but he was assuming factories on the Moon! A planet chauvinist! I think of LSP and L5 station as fairly equal prospects, but yes Moon vacuum/sun plus aim at Earth plus stable plus no junk plus no light pollution plus means that the location of the resources is NOT why I like it. In fact, I think the stuff *perhaps* will be made in space even if from lunar material as it has to spread all over, perhaps easier from orbit. Perhaps! Depends a lot on how much micr0g is being done otherwise. How much material is being captured from Bennu, or launched from Moon.

    "Why not use the material where it already is located?" If you are talking about using an asteroid for material that is in Space, of course! If you are talking setting up in a gravity well to use the material that forms the well, think micr0g, for starters. O'Neill has all in his book "The High Frontier".

48. While in orbit has lots of desirable attributes for doing things, vacuum IS kind of defined by a lack of material, and having material IS sort of necessary to most activities.

    Thus the question immediately arises when planning on doing something in space, "Why not use the material where it already is located?"

    Seriously, is this not a good deal of the argument for your favorite version of space solar power, Criswell? "Why not use the lunar material on the Moon, instead of shipping it to geosynchronous orbit?"

    I'm quite sure Musk can be persuaded to do in orbit things that sound engineering studies say are more cost effective in orbit. That's not building colonies, yet.

49. One nice thing about hordes of robots learning to cope with the world is that, if even one of them gets it figured out, you can just copy their neural network.

50. Sats that are in a position relative to Earth as their main attribute are not really even in Space, except by accident. If comm sats worked better 10 ft below ground, that is where they would be. Everything that happens in micr0g (as opposed to the physical constant G) for a reason associated with micr0g (such as it is impossible elsewhere) will not be done by Musk. He is against the idea of being in orbit! He can't understand why anyone would want to be there. Really!

    It is for this reason that I do not consider launched Solar Power Sats to be a "Space" project. Building them from non launched resources, definitely a Space project. Doing experiments and production in the ISS is a Space project. See what they say about micr0g during the calls on NASA TV. Forget planets.

51. One of the reasons it's a catastrophe, especially if we don't have major breakthroughs in gerontology, and soon, is elder care. Humanoid robots might be good for that.

52. Yes, everything that happens in micro G will be done by somebody other than the guy who owns the majority of satellites in orbit. Obviously. (◔_◔)

53. Setting aside that there must have been some weird politics going on to prompt the psychiatric community to decide to call Aspergers a type of "autism", (Let's face it, the DSM is a political document, has been for decades.) I find it interesting that it's just about the only immutable characteristic you're still allowed to use as an insult these days. It's even gaining popularity as an insult, on the left. Maybe because liberal arts majors get to decide who it's OK to insult, and Aspergers is mostly found in engineering majors, so it's never going to be thrown in their own faces? Or just because the left has taken so many objectively stupid stances, that they're trying to make objectivity look bad?

    Anyway, that's not to say that AGI isn't fairly dangerous if it ever starts to approach human level. Right now Tesla is struggling to achieve chipmunk level performance in a restricted domain, driving on roads.

54. AGI is a bad idea . I don't understand why you nerds are so excited about it . Elon is an autistic narcissist man ( human robot ) .

55. Elon sees depopulation as a major imminent global catastrophe. If robots can take the place of humans a lot of that problem goes away.

56. Nice article, a lot of useful info!
    _________________________________

    https://www.siulaudarba.lt/

57. Tesla has already demonstrated some ability to guess what is on a human mind in getting the car to anticipate the intentions of other drivers on the road. More work to be done, I'd agree. But it doesn't seem impossible in principle to make the car recognize the signals of someone directing traffic and whether they are likely legitimate (police uniform, safety vest and hard hat, hand-held STOP/GO sign, providing directions that wouldn't lead to an accident, etc).

58. Of course, everything that happens in micr0g will be done by others than Musk. And that is the future!

59. rabbit hole

60. Here is a test question for AGI: "Is the surface of a planet the right place for an expanding technological civilization?"

61. Maybe Tesla has some breakthrough AI tricks up its sleeve. They are clever in how they do things.  

    But when it comes to published AI research, Google/Deepmind, Microsoft, Open AI, Amazon and Facebook are all pretty far ahead of the pack. They have the largest and deepest AI R&D labs. But again, maybe Tesla will somehow show them all how it’s done.

62. Pure AI will never work as good as hybrid AI combined with a lot of traditional programming and additional guidelines. The type of AGI that the Musk is focusing on will have some applications but will move forward at a slower pace, always staying at the sidelines.

63. What do you mean by "rabbit track"? Doesn't seem to be a common idiom, certainly not familiar to me. "Like the Roadster" didn't clarify it for me much.
    Do you mean it's a very early predecessor of eventual "real" products, but not to be taken seriously for some time yet?

64. "I am not seeing the need for the push for androids." 

    Could they be going the embodied/enactive route to deal with the ToM (theory of the mind) and the common sense issues raised in your previous post on L5 driving? Build hordes of humanoid Telsa bots with artificial curiosity and fear in order to drive them to explore and attend to their environment and learn common sense about people and things some what like children in the process.
    https://www.youtube.com/watch?v=Jx6-DKXgAKU

65. most of these overlap with a notional venn circle that represents the neglected potential of legislation. WRT what AI drivers and logistics (e.g. toxic spill notices) have to offer.

    A car cannot run a red light, the same as a pedestrian can't walk their own 'red light' .
    AI drivers being accommodated likewise would improve 'autopilot' and human safety stats.

66. It seems Tesla, SpaceX et al are going ahead of the pack more and more, every time with the starting advantage of several billion dollars worth of previous technology investments. Like building an onion of gigantic proportions, layer after layer, where every new layer is only possible because of the previous one.

    There are a few that can still compete in some fronts, but the combined capability cluster of space launchers, communication sats, FSD cars, battery, AI and robotics nobody but them has it.

    It's kind of weird. Seems we are hitting some new meta-phase of human economic and technological development, when such advances are only possible because we have these gigantic companies, and also some loony CEO who wants to use all that to go to space and make people live on Mars.

    Said CEO is a bit of P.T. Barnum and a bit of Nikola Tesla, but it's definitely impacting the pace of development big way. Hopefully he doesn't end up as Howard Hughes.

67. Yep. A lot of the corner cases of driving are those of interpretation of events happening outside the car's owner itinerary.

    They got a hang on several already, like red lights, repair signals, bumps, suddenly crossing pedestrians, etc.

    But the weirder ones are indeed those coming from people doing weird unexpected things. Humans are rich in foolishness and weirdness.

68. I have high hopes for the Starlink system allowing the free flow of information world-wide.

69. Yes, they did have to change their approach. Still, unclear how they are going to make it able to roll with the punches when weird things happen that involve people outside the car directing them, like police, construction workers and crossing guards. And telling the difference between them and some kid waving his hands around talking on his cellphone.

70. Existing ML architectures are insufficient for L5, a properly implemented L4 system is workable in controlled environments. Waymo represent the current state of the art L4, their driverless taxi service has been in operation for a few years. That kind of result is achievable by any one with a bit of resources. Tesla spent too many years chasing deadend architectures that never had an evolutionary pathway to an autonomous system.

71. I think yes, the push for self driving will develop most of the core technology for Tesla Bot. Once you can 'see', the rest seems pretty easy. We'll see how 'intelligent' they are but as long as they can do the dishes and the laundry.. maybe the back yard.. I'd probably buy one. But this neuralink stuff he should probably give up on for a while. Hyperloops. Sure. Space networks and trips to Mars.. sure. go for it.

72. I am not seeing the need for the push for androids. I think this is a rabbit track, just like the Roadster.
    Focus should be on the Semi, the Cybertruck, an eVTOL, and at SpaceX, technology to live in space indefinitely regardless of resupply. And making that solution as compact and low mass as possible.
    Also building the first artificial gravity space station.

73. Full self driving cannot truly exist so long as the programs just react to the road and motions they see. They need to understand the minds of those who direct traffic, and understand what they are saying.
    Real-time images from satellites, and staying informed on traffic conditions could help. It would then not have to be told that the road is washed out, or there is some toxic spill, or whatever.

Comments are closed.