Lex Fridman Podcast 12 novembre 2019 Deuxième passage d'Elon Musk chez Lex Fridman (« Neuralink, IA, Autopilot et le Pale Blue Dot ») : interfaces cerveau-machine, conduite autonome, intelligence artificielle et plaidoyer pour rendre l'humanité multiplanétaire.
Deuxième passage d'Elon Musk chez Lex Fridman (« Neuralink, IA, Autopilot et le Pale Blue Dot ») : interfaces cerveau-machine, conduite autonome, intelligence artificielle et plaidoyer pour rendre l'humanité multiplanétaire.
Transcription Speaker A
The following is a conversation with Elon Musk Part 2 the second time we spoke on the podcast with parallels if not in quality than an outfit to the objectively speaking greatest sequel of all time, Godfather Part 2. As many people know, Elon Musk is a leader of Tesla, SpaceX, Neuralink and the Boring Company. What may be less known is is that he's a world class engineer and designer, constantly emphasizing first principles, thinking and taking on big engineering problems that many before him will consider impossible. As scientists and engineers, most of us
Speaker B
don't question the way things are done.
Speaker A
We simply follow the momentum of the crowd. But revolutionary ideas that change the world on the small and large scales happen when you return to the fundamentals and ask, is there a better way? This conversation focuses on the incredible engineering and innovation done in brain computer interfaces at Neuralink. This work promises to help treat neurobiological diseases, to help us further understand the connection between the individual neuron to the high level function of the human brain, and finally to one day expand the capacity of the brain through two way communication with computational devices, the Internet and artificial intelligence systems. This is the Artificial Intelligence podcast. If you enjoy it, subscribe on YouTube, Apple Podcasts, Spotify support on Patreon or simply connect with me on Twitter lexfriedman, spelled F R I D M A N. And now, as an anonymous YouTube commenter, refer to our previous conversation as the quote, historical first video of two robots conversing without supervision. Here's the second time, the second conversation with Elon Musk.
Speaker B
Let's start with an easy question about consciousness. In your view, is consciousness something that's unique to humans, or is it something that permeates all matter, almost like a fundamental force of physics?
Speaker C
I don't think consciousness permeates all matter.
Speaker B
Panpsychists believe that. Yeah, there's a philosophical how would you tell. That's true. That's a good point.
Speaker C
I believe in the scientific method or blow your mind or anything. But the scientific method is like if you cannot test the hypothesis, then you cannot reach meaningful conclusion that it is true.
Speaker B
Do you think consciousness, understanding consciousness is within the reach of science, of the scientific method?
Speaker C
We can dramatically improve our understanding of consciousness. You know, I would be hard pressed to say that we understand anything with complete accuracy. But can we dramatically improve our understanding of consciousness? I believe the answer is yes.
Speaker B
Does an AI system, in your view, have to have consciousness in order to achieve human level or superhuman level intelligence? Does it need to have some of these human qualities like consciousness, Maybe a body, maybe a Fear of mortality, capacity to love, those kinds of silly human things.
Speaker C
There's the scientific method, which I very much believe in, where something is true to the degree that it is testably so. And otherwise you're really just talking about preferences or, well, untestable beliefs or that kind of thing. So it ends up being somewhat of a semantic question where we are conflating a lot of things with the word intelligence. If we parse them out and say, are we headed towards the future where an AI will be able to outthink us in every way, then the answer is unequivocally yes.
Speaker B
In order for an AI system that needs to outthink us in every way, it also needs to have a capacity to have consciousness, self awareness, and understanding.
Speaker C
They will be self aware. Yes. That's different from consciousness. I mean, to me, in terms of what consciousness feels like. It feels like consciousness is in a different dimension, but this could be just an illusion. If you damage your brain in some way physically, you damage your consciousness, which implies that consciousness is a physical phenomenon. In my view, the things that I think are really quite likely is that digital intelligence will be able to outthink us in every way, and it will soon be able to simulate what we consider consciousness. So to a degree that you would not be able to tell the difference.
Speaker B
And from the aspect of the scientific method, it might as well be consciousness if we can simulate it perfectly.
Speaker C
If you can't tell the difference, and this is sort of the Turing Test, but think of a more sort of advanced version of the Turing Test. If you're talking to digital superintelligence and can't tell if that is a computer or a human, like, let's say you're just having a conversation over a phone or a video conference or something where you think you're talking, looks like a person, makes all of the right inflections and movements and all the small subtleties that constitute human, and talks like human, makes mistakes like a human, and you literally just can't tell, is this. Are you video conferencing with a person or an AI
Speaker B
Might as well be human. So on a darker topic, you've expressed serious concern about existential threats of AI. It's perhaps one of the greatest challenges our civilization faces. But since I would say we're kind of an optimistic descendants of apes, perhaps we can find several paths of escaping the harm of AI. So if I can give you three options, maybe you can comment. Which do you think is the most promising? So one is scaling up efforts on AI safety and beneficial AI research in hope of finding an algorithmic or maybe a policy solution. Two is becoming a multi planetary species as quickly as possible. And three is merging with AI and riding the wave of that increasing intelligence as it continuously improves. What do you think is most promising, most interesting as a civilization that we should invest in?
Speaker C
I think there's a tremendous amount of investment going on in AI. Where there's a lack of investment is in AI safety. And there should be, in my view, a government agency that oversees anything related to AI to confirm that it does not represent a public safety risk. Just as there is regulatory authority for, just like the Food and Drug Administration, there's NHTSA for automotive safety, there's the FAA for aircraft safety. We generally come to the conclusion that it is important to have a government referee or referee that is serving the public interest and ensuring that things are safe when there's a potential danger to the public. I would argue that every AI is unequivocally something that has potential to be dangerous to the public and therefore should have a regulatory agency, just as other things that are dangerous to the public have a regulatory agency. But let me tell you, the problem with this is that the government moves very slowly. And usually the way a regulatory agency comes into being is that something terrible happens. There's a huge public outcry, and years after that there's a regulatory agency or a rule put in place. Take something like seat belts. It was known for a decade or more that seatbelts would have a massive impact on safety and save so many lives and serious injuries. And the car industry fought the requirement to put seat belts in tooth and nail. That's crazy. And hundreds of thousands of people probably died because of that. And they said people wouldn't buy cars if they had seat belts, which is obviously absurd. Or look at the tobacco industry and how long they fought anything about smoking. That's part of why I helped make that movie. Thank you for smoking. You can sort of see just how pernicious it can be when you have these companies effectively achieve regulatory capture of government. The bad people in the AI community refer to the advent of digital superintelligence as a singularity. That is not to say that it is good or bad, but that it is very difficult to predict what will happen after that point and that there's some probability it will be bad, some probability it will be good. We obviously want to affect that probability and have it be more good than bad.
Speaker B
Well, let me on the merger with AI question and the incredible work that's being done at Neuralink. There's a lot of fascinating innovation here across different disciplines going on. So the flexible wires, the robotic sewing machine and responsive brain movement, everything around, ensuring safety and so on. So we currently understand very little about the human brain. Do you also hope that the work at neuralink will help us understand more about our. About the human mind, about the brain?
Speaker C
Yeah, I think the work in Neuralink will definitely shed a lot of insight into how the brain, the mind works. Right now, just the data we have regarding how the brain works is very limited. We've got FMRI that's kind of like putting a stethoscope on the outside of a factory wall and then putting it all over the factory wall. And you can sort of hear the sounds, but you don't know what the machines are doing really. It's hard. You can infer a few things, but it's very broad brushstroke. In order to really know what's going on in the brain, you really need. You have to have high precision sensors. And then you want to have stimulus and response. Like if you trigger a neuron, how do you feel? What do you see? How does it change your perception of
Speaker B
the world you're speaking to? Physically, just getting close to the brain, being able to measure signals from the brain will give us sort of open the door inside the factory.
Speaker C
Yes, exactly. Being able to have high precision sensors that tell you what individual neurons are doing, and then being able to trigger a neuron and see what the response is in the brain. So you can see the consequences of if you fire this neuron, what happens? How do you feel? What does change? It'll be really profound to have this in people because people can articulate their change. Like if there's a change in mood or if they can tell you if they can see better or hear better, or be able to form sentences better or worse, or their memories are jogged or that kind of thing.
Speaker B
So on the human side, there's this incredible general malleability plasticity of the human brain. The human brain adapts, adjusts, and so on.
Speaker C
So that's not that plastic, to be totally frank.
Speaker B
So there's a firm structure, but nevertheless there's some plasticity. And the open question is, so if I could ask a broad question, is how much that plasticity can be utilized, sort of on the human side, there's some plasticity in the human brain. And on the machine side, we have neural networks, machine learning, artificial intelligence. It's able to adjust and figure out signals. So there's mysterious language that we don't perfectly understand that's within the human brain. And then we're trying to understand that language to communicate both directions. So the brain is adjusting a little bit. We don't know how much, and the machine is adjusting. Where do you see as they try to sort of reach together, almost like with an alien species, Try to find a protocol, communication protocol that works. Where do you see the biggest. The biggest benefit arriving from? On the machine side or the human side? Do you see both of them working together?
Speaker C
I think the machine side is far more malleable than the biological side by a huge amount. So it will be the machine that adapts to the brain. That's the only thing that's possible. The brain can't adapt that well to the machine. You can't have neurons start to regard an electrode as another neuron, because neurons just. This is like the pulse, and so something else is pulsing. So there is that elasticity in the interface, which we believe is something that can happen. But the vast majority of the malleability will have to be on the machine side.
Speaker B
But it's interesting, when you look at that synaptic plasticity at the interface side, there might be like an emergent plasticity because it's a whole nother. It's not like in the brain. It's a whole nother extension of the brain. You know, we might have to redefine what it means to be malleable for the brain. So maybe the brain is able to adjust to external interfaces.
Speaker C
There will be some adjustments to the brain because there's going to be something reading and simulating the brain, and so it will adjust to that thing. But the vast majority of the adjustment will be on the machine side. If this is just. It has to be that, otherwise it will not work. Ultimately, we currently operate on two layers. We have sort of a limbic prime primitive brain layer, which is where all of our kind of impulses are coming from. It's sort of like we've got like a monkey brain with a computer stuck on it. That's the human brain. And a lot of our impulses and everything are driven by the monkey brain and the computer. The cortex is constantly trying to make the monkey brain happy. It's not the cortex that's steering the monkey brains, the monkey brain steering the cortex, you know, but the cortex is
Speaker B
the part that tells the story of the whole thing. So we convince ourselves it's. It's more interesting than just the monkey brain.
Speaker C
Cortex is like what we call, like, human intelligence. You know, so it's like, that's like the advanced computer relative to other creatures. Other creatures do not have either really. They don't have the computer or they have a very weak computer relative to humans. But it sort of seems like surely the really smart thing should control the dumb thing, but actually the dumb thing controls the smart thing.
Speaker B
So do you think some of the same kind of machine learning methods, whether that's natural language processing applications, are going to be applied for of the communication between the machine and the brain to learn how to do certain things like movement of the body, how to process visual stimuli and so on? Do you see the value of using machine learning to understand the language of the two way communication with the brain?
Speaker C
Sure, yeah, absolutely. I mean we're a neural net and that, you know, AI is basically neural net. So it's like digital neural net will interface with biological neural net and hopefully bring us along for the ride. But the vast majority of our intelligence will be digital. Think of the difference in intelligence between the cortex and your limbic system is gigantic. Your limbic system really has no comprehension of what the hell the cortex is doing. It's just literally hungry or tired or angry or sexy or something. And then that communicates that impulse to the cortex and tells the cortex to go satisfy that. Then a great deal of a massive amount of thinking, a truly stupendous amount of thinking has gone into sex without purpose, without procreation. Without procreation, which. Which is actually quite a silly action. In the absence of procreation, it's a bit silly. So why are you doing it? Because it makes the limbic system happy. That's why.
Speaker C
But it's pretty absurd really.
Speaker B
Well, the whole of existence is pretty absurd in some kind of sense.
Speaker C
Yeah, but I mean this is a lot of computation has gone into. How can I do more of that with procreation not even being a factor? This is I think a very important area of research by nsfw,
Speaker B
an agency that should receive a lot of funding. Especially after this conversation.
Speaker C
I propose the formation of a new agency.
Speaker B
Oh boy. What is the most exciting or some of the most exciting things that you see in the future impact of neuralink both on the science, engineering and societal broad impact.
Speaker C
So neuralink I think at first will solve a lot of brain related diseases. So create anything from like autism, schizophrenia, memory loss. Like everyone experiences memory loss at certain point in age, parents can't remember their kids names and that kind of thing. So there's I think a tremendous amount of Good that neuralink can do in solving. Critical damage to the brain or the spinal cord. There's a lot that can be done to improve quality of life of individuals, and that will be. Those will be steps along the way and then ultimately is intended to address the risk, the existential risk associated with digital superintelligence, like we will not be able to be smarter than a digital supercomputer. So therefore, if you cannot beat them, join them and release. We won't have that option.
Speaker B
So you have hope that neuralink will be able to be a kind of connection to allow us to merge, to ride the wave of the improving AI systems.
Speaker C
I think the chance is above zero percent.
Speaker B
So it's non zero. There's a chance. And that's.
Speaker C
Have you seen Dumb and Dumber?
Speaker C
So I'm saying there's a chance.
Speaker B
You're saying one in a billion or one in a million, whatever it was.
Speaker C
And Dumb and Dumber, you know, it went from maybe one in a million to improving. Maybe it'll be one in 1,000 and then one in 100, then one in 10. Depends on the rate of improvement of neuralink and how fast we're able to make progress, you know.
Speaker B
Well, I've talked to a few folks here, they're quite brilliant engineers, so I'm excited.
Speaker C
Yeah, I think it's like, fundamentally good, you know, giving somebody back full motor control after they've had a spinal cord injury, you know, restoring brain functionality after a stroke, solving debilitating genetically orange brain diseases. These are all incredibly great, I think. And in order to do these, you have to be able to interface with neurons at a detailed level and you build fire the right neurons, read the right neurons, and then effectively you can create a circuit, replace what's broken with silicon, and essentially fill in the missing functionality. And then over time, we can have, we develop a tertiary layer. So if, like, the limbic system is a primary layer, then the cortex is like the second layer. And I said, obviously the cortex is vastly more intelligent than the limbic system. But people generally like the fact that they have a limbic system and a cortex. I haven't met anyone who wants to delete either one of them. They're like, okay, I'll keep them both. That's cool.
Speaker B
The limbic system is kind of fun.
Speaker C
That's where the fun is.
Speaker C
And then people generally don't want to lose their cortex either. So, like having the cortex and the limbic system. Yeah, and. And then there's a tertiary layer which will be digital superintelligence and I think there's room for optimism given that the cortex, the cortex is very intelligent and limbic system is not, and yet they work together well. Perhaps there can be a tertiary layer where digital superintelligence lies and that will be vastly more intelligent than the cortex, but still coexist peacefully and in a benign manner with the cortex and limbic system.
Speaker B
That's a super exciting future both in low level engineering that I saw as being done here, and the actual possibility in the next few decades.
Speaker C
It's important that neuralink solve this problem sooner rather than later, because the point at which we have digital superintelligence, that's when we pass singularity and things become just very uncertain. It doesn't mean that they're necessarily bad or good, but the point at which we pass singularity, things become extremely unstable. So we want to have a human brain interface before the singularity, or at least not long after it, to minimize existential risk for humanity and consciousness as we know it.
Speaker B
But there's a lot of fascinating actual engineering low level problems here at neuralink that are quite exciting.
Speaker C
The problems that we face neuralink are material science, electrical engineering, software, mechanical engineering, microfabrication. It's a bunch of engineering disciplines. Essentially that's what it comes down to is that you have to have a tiny electrode. It's so small it doesn't hurt neurons, but it's got to last for as long as a person, so it's going to last for decades. And then you've got to take that signal, you've got to process that signal locally at low power. So we need a lot of chip design engineers because we're going to do signal processing and do so in a very power efficient way so that we don't heat your brain up, because the brain's very heat sensitive. And then we've got to take those signals and we're going to do something with them, and then we've got to stimulate back so you could bidirectional communication. So if somebody's good at material science, software, mechanical engineering, electrical engineering, chip design, microfabrication, those are the things we need to work on. We need to get good at material science so that we can have tiny electrodes that last a long time. And it's a tough thing with the material science problem, is a tough one, because you're trying to read and simulate electrically in an electrically active area. Your brain is very electrically active and electrochemically active. So how do you have, say, a coating on the electrode that doesn't dissolve over time and is safe in the brain. This is a very hard problem. And then how do you collect those signals in a way that is most efficient because you really just have very tiny amounts of power to process those signals and then we need to automate the whole thing. So it's like Lasik. If this is done by neurosurgeons, there's no way it can scale to large numbers of people. And it needs to scale to large numbers of people because I think ultimately we want the future to be determined by a large number of humans.
Speaker B
Do you think this has a chance to revolutionize surgery, period? So neurosurgery and yeah, for sure, it's
Speaker C
got to be like Lasik. If LASIK had to be hand done, done by hand by a person, that wouldn't be great, you know, so it's done by a robot and the ophthalmologist kind of just needs to make sure your head's in the right position and then they just press a button and go.
Speaker B
So smart summon. And soon autopark takes on the full beautiful mess of parking lots and their human to human non verbal communication. I think it has actually the potential to have a profound impact in changing how our civilization looks at AI and robotics. Because this is the first time human beings, people that don't own a Tesla, may have never seen a Tesla or heard about a Tesla, get to watch hundreds of thousands of cars without a driver.
Speaker B
Do you see it this way? Almost like an education tool for the world about AI. Do you feel the burden of that, the excitement of that, or do you just think it's a smart parking feature?
Speaker C
I do think you are getting at something important, which is most people have never really seen a robot. And what is the car that is autonomous? It's a four wheeled robot, right?
Speaker B
Yeah. It communicates a certain sort of message with everything from safety to the possibility of what AI could bring to its current limitations, its current challenges. It's what's possible. Do you feel the burden of that? Almost like a communicator educator to the world about AI.
Speaker C
We're just really trying to make people's lives easier with autonomy. But now that you mention it, I think it will be an eye opener to people about robotics because they've really never seen most people never seen a robot. And there are hundreds of thousands of Teslas. Won't be long before there's a million of them that have autonomous capability and the drive without a person in it. And you can see the kind of evolution of the car's personality and thinking. With each iteration of autopilot, you can see it's uncertain about this or it gets. But now it's more certain. Now it's moving in a slightly different way. I can tell immediately if a car is on Tesla Autopilot because it's got just little nuances of movement. It just moves in a slightly different way. Cars on Tesla Autopilot, for example, on the highway are far more precise about being in the center of the lane than a person. If you drive down the highway and look at how at where cars are, the human driven cars are within their lane. They're like bumper cars. They're like moving all over the place. The car on autopilot, dead center.
Speaker B
Yeah. So the incredible work that's going into that neural network, it's learning fast. Autonomy is still very, very hard. We don't actually know how hard it is fully. Of course, you look at most problems, you tackle this one included with an exponential lens. But even with an exponential improvement, things can take longer than expected sometimes. So where does Tesla currently stand on its quest for full autonomy? What's your sense? When can we see successful deployment of full autonomy?
Speaker C
Well, on the highway already, the probability of intervention is extremely low.
Speaker C
So for highway autonomy, with the latest release especially, the probability of needing to intervene is really quite low. In fact, I'd say for stop and go traffic, it's far safer than a person. Right now. The probability of an injury or an impact is much, much lower for the pilot in a person. And then with navigate and autopilot, you can change lanes, take highway interchanges, and then we're coming at it from the other direction, which is low speed, full autonomy. And in a way, this is like, how does a person learn to drive? You learn to drive in the parking lot. First time you learn to drive. Probably wasn't jumping on Market street in San Francisco. That'd be crazy. You learn to drive in the parking lot, get things right at low speed, and then the missing piece that we're working on is traffic lights and stop streets, soft streets, I would say actually also relatively easy because you kind of know where the stop street is. Worst case, you can geocode it and then use visualization to see where the line is and stop at the line to eliminate the GPS error. So actually I'd say it's probably complex. Traffic lights and very windy roads are the two things that need to get solved.
Speaker B
What's harder? Perception of control for these problems. So being able to perfectly perceive everything or figuring out a plan once you perceive everything. How to interact with all the agents in the environment. In your sense, from a learning perspective, is perception or action harder in that giant, beautiful multitask learning neural network?
Speaker C
The hardest thing is having accurate representation of the physical objects in vector space. So taking the visual input, primarily visual input, some sonar and radar, and then creating an accurate vector space representation of the objects around you. Once you have an accurate vector space representation, the planning and control is relatively easier. That says relatively easy. Basically, once you have accurate vector space representation, then you're kind of like a video game, like cars in Grand Theft Auto or something. They work pretty well. They drive down the road, they don't crash pretty much unless you crash into them. That's because they've got an accurate vector space representation of where the cars are and then they're rendering that as the output.
Speaker B
Do you have a sense high level that Tesla's on track on being able to achieve full autonomy? So on the highway, yeah, absolutely. And still no driver state, driver sensing.
Speaker C
And we have driver sensing with torque on the wheel.
Speaker B
That's right, yeah. By the way, just a quick comment on karaoke. Most people think it's fun, but I also think it is a driving feature. I've been saying for a long time, singing in the car is really good for attention management and vigilance management.
Speaker C
Tesla karaoke is great. It's one of the most fun features of the car.
Speaker B
Do you think of a connection between fun and safety? Sometimes, yeah.
Speaker C
If you can do both at the same time, that's great.
Speaker B
I just met with Andrew N. Wife of Carl Sagan, who directed Cosmos.
Speaker C
I'm generally a big fan of Carl Sagan. He's super cool and had a great way of putting things all about consciousness, all civilization, everything we've ever known and done is on this tiny blue dot. People also get too trapped in their squabbles amongst humans. Don't think of the big picture. They take civilization and our continued existence for granted. They shouldn't do that. Look at the history of civilizations. They rise and they fall. And now civilization is all. It's globalized. And so civilization, I think now rises and pulls together. There's not geographic isolation. This is a big risk. Things don't always go up. That should be. That's an important lesson of history.
Speaker B
In 1990, at the request of Carl Sagan, the Voyager 1 spacecraft, which is a spacecraft that's reaching out farther than anything human made into space, turned around to take a picture of Earth from 3.7 billion miles away. And as you're talking about the pale blue dot. That picture there takes up less than a single pixel in that image.
Speaker B
Appearing as a tiny blue dot, as pale blue dot as Carl Sagan called it. So he spoke about this dot of ours in 1994. And if you could humor me, I was wondering if in the last two minutes you could read the words that he wrote described in this. Be able.
Speaker C
It's funny. The universe appears to be 13.8 billion years old. Earth is like four and a half billion years old. In another half billion years or so, the sun will expand and probably evaporate the oceans and make life impossible on Earth. Which means that if it had taken consciousness 10% longer to evolve, it would never have evolved at all, just 10% longer. And I wonder, I wonder how many dead one planet civilizations there are out there in the cosmos that never made it to the other planet and ultimately extinguished themselves or were destroyed by external factors. Probably a few. It's only just possible to travel to Mars. Just barely. FG was 10% more. Wouldn't work, really. If G was 10% lower, it would be easy. Like you can go single stage from the surface of Mars all the way to the surface of the Earth. Because Mars is 37% Earth's gravity, we need a giant booster. Get off Earth. Channeling Carl Sagan. Look again at that dot. That's here. That's home. That's us. On it. Everyone you love, everyone you know, everyone you've ever heard of, every human being who ever was lived out their lives, the aggregate of our joy and suffering. Thousands of competent religions, ideologies and economic doctrines, every hunter and forager, every hero and coward, every creator and destroyer of civilization, every king and peasant, every young couple in love, every mother and father, hopeful child, inventor and explorer, every teacher of morals, every corrupt politician, every superstar, every supreme leader, every saint and sinner in the history of our species lived there on a mote of dust suspended in a sunbeam. Our planet is a lonely speck in the great enveloping cosmic dark. In our obscurity, in all this vastness, there is no hint that help will come from elsewhere to save us from ourselves. The Earth is the only world known so far to harbor life. There is nowhere else, at least in the near future, to which our species could migrate. This is not true. This is false. Mars.
Speaker B
And I think Carl Sagan would agree with that. He couldn't even imagine it at that time. So thank you for making the world dream and thank you for talking today. I really appreciate it.