“That’s why I created xAI which is to have an AI that is maximally truth seeking, that aspirationally does love humanity and will you seek the best interests of of humanity going forward.” — ELON MUSK
Elon Musk and Dr. Peter Diamandis FII8. Conversation on the Future of AI
“When Google Gemini came out uh you people asked U whether it is more which is worse misgendering Caitlyn Jenna or Global th nuclear war and it said misgendering Caitlyn Jenna which is obviously a problem because you know we all die in global thir nuclear war and so if you have have an AI That’s programmed for things like that it could conclude that the best way to ensure that nobody is misgendered is to uh annihilate All Humans thus making the probability of a future misgendering zero you that’s highly problematic um so so you really want to have a maxim truth seeking Ai and um I can’t emphasize that enough that’s incredibly important um and obviously built an AI that loves Humanity um and um you know and I think these so so I’m a little concerned that’s why I created xai which is to to have an AI that is maximally truth seeking um that aspirationally does love humanity and will you seek the best interests of of humanity going forward. – Elon Musk, Future Investment Initiative, Oct 29, 2024
SPECIAL CONVERSATION: THE FUTURE OF AI took place at the King Abdulaziz International Conference Center in Riyadh on October 29, 2024, during the 8th edition of the Future Investment Initiative.
“Probably all countries will have their own AI clusters over time.” – Elon Musk
Introduction ladies and gentlemen prepare for a fun conversation with a very special guest and someone I’m proud to call a friend welcome Elon [Applause] Musk welcome to uh to Riad um it’s been quite an incredible 90 days for you my friend uh yeah xai Colossus came online 122 days you’re about to double in size again uh the extraordinary success of of the fifth mission of Starship uh and the booster capture amazing the launch of cyber cab starlink helping disaster relief progress with Optimus the second neurolink human patient new record users on X uh and that’s in 90 days I’m sure I missed a few things welcome yeah I’ve been playing as significant role in this election I I think I I saw you someplace on X dealing with election issues um let’s talk a little bit about AI which is on the tips of everybody’s uh conversations here uh when we spoke last in March at the abundance Summit your prediction was that AI was increasing at a rate as fast as a 100 times per year and that by 2029 or 2030 we might see AI as capable as 8 billion humans are you still seeing that pace yeah it depends it’s difficult it’s a difficult thing to quantify exactly but um I certainly feel comfortable saying that it’s getting 10 times better per year which is you know let’s say it’s you know four years from now that would mean 10,000 times better so maybe 100,000 yeah and it’s it’s it’s it I think it will be able to do anything that any human can do um possibly within with the next uh year year or two and then uh what what can how long how much longer than that does it take to be able to do what all humans can do combined I think not not long probably only I don’t know 3 years from that point so it like 20 20 29 is 28 something like that the other conversation we’ve had and you came out in the same side as Jeffrey Hinton in this was 80% probability it’s going to be awesome 20% probability was screwed are you still you still on those odds yeah I mean one can say it’s it’s it’s it’s most likely going to be great and there’s there some chance which could be 10 to 20% that um it goes bad um the chance is not zero that it goes bad but overall one one could say the cup is 80% full uh is one positive way to look at it maybe 90% I like that increasing OD what keeps you up yeah what keeps you up at night besides running six companies um well about waking up early for to participate in uh talks like this one I I felt very guilty asking you to to do this but thank you for for joining none the least but in terms of worried about the future um well I think AI a significant existential threat and something we should be pay paying close attention to uh it’s probably the most significant near-term threat the longer term than that is the global population collapse you know both rates have been collapsing pretty much worldwide um and we’re headed to you know a situation where for example based on current birth rates uh South Korea would have about a third of its current population have much less um Europe would have about half of its current current population half much less and I should say those numbers are if the birth rate were suddenly to return to St to to 2.1 per woman which is a stability Point um which is is not doing so if the if if the current compounding uh effect continues um you would see really um many countries become 5% of their current size or less within three generations I know you’ve been doing I would consider that to be a very big very big problem um and and I think if for most countries they should view the birth rate as as the single biggest problem they need to solve um I mean if you don’t make new humans there’s no humanity and and all the policies in the world don’t matter I know you’ve been doing your part to maintain the birth rate in in the us yes I am I mean you know it’s you’ve got to walk the talk um so I do have a lot of kids and I encourage others to have lots of kids but on the AI and I’m sorry to go How do you protect against the downside negative on this but what do you what do you what are you doing right now that’s most important for uh countering that 10% probability of dystopian outcomes is there something or do you is there a regulation that you’re you’re promoting how do you think about uh you know the upside will take care of itself how do you protect against the downside well my my instinct with respect to AI safety is that you have to be create a maximally truth seeking AI uh which may sound obvious but that’s what I’ve seen being produced is is not maximally truth seeking um it tends to be trained to be politically correct um and for a lot of the AIS that are being uh trained in um in the San Francisco Bay Area they are uh that they have they have taken on the philosophy of the people around them which kind of makes sense uh so you know you have sort of a woke um nihilistic in my opinion um philosopy that is being built into these AIS um and they’re being thought to say crazy things in some cases uh that are that are very troubling um like when Google Gemini came out uh you people asked U whether it is more which is worse misgendering Caitlyn Jenna or Global th nuclear war and it said misgendering Caitlyn Jenna which is obviously a problem because you know we all die in global thir nuclear war and so if you have have an AI That’s programmed for things like that it could conclude that the best way to ensure that nobody is misgendered is to uh annihilate All Humans thus making the probability of a future misgendering zero you that’s highly problematic um so so you really want to have a maxim truth seeking Ai and um I can’t emphasize that enough that’s incredibly important um and obviously built an AI that loves Humanity um and um you know and I think these so so I’m a little concerned that’s why I created xai which is to to have an AI that is maximally truth seeking um that aspirationally does love humanity and will you seek the best interests of of humanity going forward no you just tweeted that you’re doubling the size of the Colossus net um cluster um what are your thoughts we already have with XII the most powerful training cluster in the world and we’re about to devil it um energy is a a point of Energy conversation here um How concerned are you about providing sufficient energy for the growing hungry clusters globally yeah I think things things are currently uh chip limited or or they’re not quite chip limited they’re they get to the point where they’re limited by voltage Transformers and installation um and they will become limited by energy um so there will be a tremendous amount of energy that’s needed for for uh digital intelligence and for um and for also for electrification of Transport so those two things are a big deal um yeah we’re going to need a lot of energy the longterm almost all the energy that we will get is going to come from the Sun um so one way to look at civilization is progress on the cter shf scale we’re just barely getting to one well we’re far from I think we’re probably we might be close we might I’m not it’s not clear to me we’re above 1% on the qu scale one cuz K scale one means you’ve you’ve harnessed all the power of a planet I think we I think we’ve probably harness less than 1% of the power of Earth um our CIP scale to is you’ve harest all the power of your son um the sun is overwhelmingly the the largest source of energy in the solar system everything else is maybe amounts to about a trillion of the energy in the solar system compared to the Sun less go safely less less than a trillionth of all the energy is nons sun in our solar system yeah we’re using 1 8,000th of the sun’s energy hits your surface the Earth just that just that hits the surface of the Earth yeah yes and the the percentage of the sun’s energy that hits the surf surface of the Earth is um is less than a trth of the energy that the sun produces so um almost all energy longterm will be solar call it round it rounds up to 100% so it rounds up to 100% that that’s uh how much of the energy in the future will be solar um when you when you view things from a quad ship standpoint you know you we have a number of national AI Transformation leaders uh corporate Financial leaders from the Middle East here what’s your advice to decision makers here in the room that don’t want to miss the AI transformation that will be part of the leadership of that AI transformation do they need to build their own clusters here are they partnering yeah I I well I I think there probably all countries will have their own AI clusters over time um it’s currently very difficult to actually build an AI cluster and have it run um that it’s not like just pulling a computer out of a box it they they are currently very difficult to run um and you have to say oh you going to be training a Frontier Model um because if you’re training a Frontier Model then you you need a massive amount of compute and a level of technical skill that is only a few a few companies POS possess um so so but but over time I think every country will have uh AI compute clusters um it’s just it’s just going to be a normal thing that every country has so yeah um so basic infrastructure for every nation like they have an electrical grid yeah it’ll be something like electricity or you know uh just or or you know having an airline or something like that it’s every every country will have AIS or multiple AIS so um and there will be a lot of robots there’ll be a lot of robots but we had way more robots Than People yeah let’s Humanoid Robots have that conversation a second because people are concerned about uh as you said dwindling populations Ai and and robots have potential to help support the GDP um congratulations on Optimus 2 and soon Optimus 3 uh your prediction on the number of robots by 2040 humanoid robots to be specific what order of magnitude by 2040 yeah so um I think by if you say like 2040 probably there are more humanoid robots than ra people so 10 billion yeah yes and your price point on these humanoid robots you’re you’re pretty good on pricing some sometimes you’re off on timing yeah I’m I’m often optimistic on timing but um although you know the Press will report when I’m late but not when I’m early um you know for example our Shanghai Factory uh we thought it would take about a year and a half and we did it in 11 months um our gavada Factory we thought 2 years we did it in 18 months um or the Colossus cluster Texas Factory two years we did it in 14 months so I’ve been early actually many times it’s just it’s just not reported um so when I when I make a prediction I I try to figure out I try to say what what is the 50th percentile likely which means that half the time I should be wrong um so I’m not sand backing essentially um um so but but but I I think that’s once you get out 240 that’s a long time from now um going 25 years there’ll be at least 10 billion humanoid robots um and price price yeah the price point will be I think quite low um probably 20 $25,000 for a robot that could do anything um we will be in a future in assuming we are in the good path of AI I think we will be in a future of abundance you know obviously you wrote a book called abundance so I think you would agree that that is probably the outcome um that that uh basically anyone will anyone will be able to have any goods and services they want the the actual marginal cost of goods and services will be extremely low in the future let me so in our last four minutes let me change the subject Starship on Mars subject to something near and dear to both of our hearts uh congratulations on on Starship uh it was literally awesome probably the engineering feat of this decade if not more not bad for humans not bad for hum we did that with no AI was involved in that whatsoever so um amazing I’m I’m glad to say that we’re a that entirely with human brains and without AI I think in the future the AI might look at that and say not bad for a bunch of monkeys when are we on Mars when are when is uh is Starship on Mars I think I think we’ll be we’ll be able to launch some Starships to Mars in two years um so at the nor the next Mars window which is an about 26 27 months but we’re just about to start uh or we’re we’re at the beginning of of a mar Transit window now and they occur every 26 months so um just over two years we’ll be sending our first uncrewed Starship suar and then if those if those work out well and we don’t increment the crater count Ms then we’ll send humans uh two years after that so a challenge to be on the surface of Mars before the end of the decade sounds like a a reasonable um yeah Proclamation for either for for either side of the White House hopefully yeah I feel more optimistic about it under a with a trump White House than a non-trump white house because um the biggest impediment progress that we’re experiencing is uh regulatory um is is overregulation got to keep those whales and sharks safe yeah yeah exactly um it it just takes us uh I mean it’s take takes longer to get the permit to launch than to build a giant rocket and the the bureaucracy in the US has has been growing every year and has particularly grown under the bid Administration um and uh unless we do something to scale back the uh overregulation I call it sort of um America is getting and a lot of countries are getting slow strangulation from overregulation unless something is done to push back on that um it’ll eventually become illegal to do almost any large project and uh we won’t be able to get tomorrow last subject um congrats on the Cyber cab rollout um uh pretty extraordinary Kathy Woods predicted it to be you know multiple trillions of dollars of potential uh uh GDP growth and impact um yeah give us give us some predictions on on when we’ll see cyber cab when we’ll be ordering a cyber cab yeah well Tesla um unsupervised full self-driving we expect to be working in the US next year with the the model 3 and Y so you you don’t have to wait for the Roo taxi or cyoc cap to uh for for Tes to release autonomy we’re currently expecting to exceed human safety levels um in Q2 next year um and then substantially go beyond that thereafter and and so really it’s just a software update to the cars um to be able to uh do launch our self-driving Network we so we expect to unsupervised full self driving and and start that in California and uh Texas you know around the middle of next year um and then at you know we have 7 million cars on the road uh we’ll have I 9 a half million cars by the end of next year so uh and eventually will have a fleet of I 100 million plus Vehicles U and they’ll all be autonomous um the the sa cap with no steering wheel low pedals uh we’re expecting to reach volume production in 2026 so that’s um that’s certainly interesting but like I said the the actual launch of of of a robotic taxi on two by full self driving is actually next year um and at the at the event the Tesla autonomy event so we had 50 cars 30 model wise that were driverless and 20 of the Cyber cabs and so autonomy is here um is what I’m saying and uh yeah and all cars will drive themselves this this is a no-brainer um they and they’ll get to where they’re 10 times saf safer than human driven cars uh which will save I don’t know past A Million Lives uh a year globally um and then Optimus uh thoughts limited production at next year 2025 and then uh should be in volume production in 26 um and then we’ll grow to I think ultimately be the biggest product uh of any kind ever um so and I kind of see I kind of agree with with ar invest and Cath would that uh autonomy like sort of robotic taxes makes Tesla kind of like a about a $5 trillion company um the Optimus robot I think makes Tesla a 25 trillion company it’s not even clear what money means in that in future we end up in a post capitalist Society at some point Elon you make we do sort of end up it it does become kind of post capitalist um and like I said and I know you agree with this that and looking at the mo the most likely bright side we’re headed for an age of abundance where anyone can have any goods and services they want it won’t be a case of universal basic income it’ll be a case of universal High income is the most likely outcome you make it look easy my friend thank you for making some time available I know it isn’t easy let’s give it up for Elon Musk everybody thanks thank you my friend