In September 2016, Sam Altman, now CEO of OpenAI, father of ChatGPT, and president of venture capital firm Y Combinator, interviewed Elon Musk at Tesla's Fremont, California factory. Musk elaborated on what created OpenAI and, for him, the five things that matter most to the future.
It was a conversation between the two co-founders of OpenAI, which, according to one YouTuber, was like "Kobe Bryant interviewed Jordan before becoming famous."
Time background: Musk's relationship with OpenAI
December 2015
Sam Altman and Elon Musk, Amazon AWS, Infosys, and YC Research announced the founding of OpenAI and injected $1 billion.
September 2016
This interview. Sam Altman was president of venture capital firm Y Combinator. Tesla just released a new Model 3 model in April and is building an automatic production line, and Lao Ma is about to fall into "production capacity hell".
In 2018
Musk withdrew from OpenAI's board seat, claiming that there may be a "potential conflict of interest" because he also serves as Tesla's CEO to develop AI-based self-driving technology.
In 2019
OpenAI transformed from a nonprofit to a "limited profit" for-profit organization.
2020/2021/2022
OpenAI has successively launched GPT-3 models, DALL-E (generating pictures through text), ChatGPT and other products.
The rubble is focused
The most important thing
- In terms of the things most likely to affect the future of humanity, I think AI is probably the most important thing in the short term.
- I think genetics is probably the second most important thing.
- 25 years ago, I thought the most important [five] things were: getting life across planets, accelerating the transition to sustainable energy, the internet in a broad sense, and then genetics and artificial intelligence.
Choose to do what works
- That's why it's great to create something that makes a huge difference, affecting a small to moderate number of people. But even if it makes a small difference, it can be just as great if it affects a large number of people.
- If people think that technology will automatically advance every year, then this is not the case. It will only progress if smart people work like crazy and try to make it better. This is the real way for any technology to advance.
- I think you should always remember that entropy is not on your side.
- I'm not without fear, my feelings are strong. But sometimes, when something is important enough and you believe in it enough, you just let it go despite your fears.
Artificial intelligence risk, the original intention of creating OpenAI
- I don't think the risk is that AI will develop its own will out of evil. The concern is that some people may use it for bad things, and even if they don't intend to use it for bad things, someone will take it away and use it for bad things.
- I think we have to democratize AI technology and make it widely available. And that's obviously why the whole team created OpenAI.
- If we can improve the brain-computer interface between the cerebral cortex and the "digitally extended you" and effectively integrate with AI, you are effectively a symbiote of AI and humans.
- If this solution is widely available and anyone can own it, then we have solved the problem of control. We don't have to worry about some evil dictatorial AI anymore, because we are AI as a whole.
Full text of the interview
Am:
Today we have Elon Musk. Elon, thank you for the interview.
Elon:
Thank you for inviting me.
Am:
Today, we want to take a moment to talk about your vision for the future and where people should be headed.
First of all, can you tell us that you famously said when you were young that you believe that there are five issues for you that are the most important work to be done. If you were back at 22 today, what are the five questions you'd want to solve?
Elon:
Well, first of all, I think if someone is doing something that works for the rest of society, that's a good thing and doesn't necessarily need to change the world.
If you're doing something that is of high value to people, frankly, even if it's just a little game, or just some kind of improvement in photo sharing, if it brings a small amount of benefit to a large group of people, I think that's fine. You don't have to change the world to be good.
But in terms of the things most likely to affect the future of humanity, I think AI is probably the most important thing in the short term.
It's important that we get AI to come in a good way, so that if you can see the future through the crystal ball, you'll be happy with that outcome, because as we've discussed many times, it can go wrong. We do need to make sure that it develops in the right way.
Artificial intelligence, working on artificial intelligence and ensuring that its future is bright. That's what I think is the most important and urgent thing right now.
And then, apparently, anything to do with genetics. If you can really solve genetic diseases, if you can prevent dementia or Alzheimer's through genetic reprogramming, that's great. I think genetics is probably the second most important thing.
I think we need to have a high-bandwidth brain interface because we currently have limited bandwidth. We have a digital "third self" in the form of emails, computers, phones, apps. In fact, we are already superhuman. But the limitations in the interface between our cerebral cortex and the third-level digital self are pretty bad.
I think it's also very important to help solve this bandwidth limitation for the future.
Vision prosthesis technology helps blind people regain partial vision. Source: Neuralink Show & Tell 2022
Am:
One of the most frequently asked questions by ambitious young people is, I want to be the next Elon Musk, how can I do this?
Obviously, the next Elon Musk will be doing something very different from you. But what of your past experiences, what did you do when you were younger, have had a huge impact on you?
Elon:
Well, first of all, I should say that I don't expect to be involved in all these projects.
In college, it was a long time coming, and 25 years ago, I thought the most important thing was: to get life across planets, to accelerate the transition to sustainable energy, the internet in general, and then genetics and artificial intelligence.
I don't think I expect to be involved in all of these projects. In fact, when I was in college, I thought that I would start by helping cars transition to electrification. In fact, the work I did during my internship was to use advanced supercapacitors to see if it would be a breakthrough for the energy storage of cars relative to batteries.
Then, when I went to Stanford, this was also my original graduate project, working on the latest energy storage technology for electric vehicles.
I put my studies on hold and started an internet company in '95. Because for some specific technologies, it did seem to be at an inflection point in the curve. And I don't want to just do a Ph.D. at Stanford and watch it happen, and I'm not entirely sure if the technology I'm working on will actually succeed. Just as you can do a Ph.D. in many directions, but it doesn't end up having a real impact on the world.
And I just want to be a useful person. This is an optimization question, what can I do to really be useful?
In 1995, Musk took a break from Stanford to start an internet company, Zip2
Am:
Do you think that if you want to be useful today, you need to read a doctorate?
Elon:
In most cases, it is not required. Some situations require, but most do not.
Am:
So, how do you become a useful person?
How should people seek ways to make themselves the most useful person?
Elon:
Whatever you want to create, what is the increase in practicality compared to the current state of the art is multiplied by the number of people affected.
That's why it's great to create something that makes a huge difference, affecting a small to moderate number of people. But even if it makes a small difference, it can be just as great if it affects a large number of people.
It's like the area under the curve. Yes, that's right, in both cases, the area under the curve is actually roughly similar. Yes, strive to be useful, influential.
Am:
Then, when you're trying to estimate the probability of success, you say, this thing is very valuable, and the area under the curve is considerable.
I want to give the example of SpaceX, when you decided to do this, at the time, it was a really crazy thing, right?
Elon:
Pretty crazy, of course. People aren't sure about it, but I agree, I agree with them, it's crazy.
If the goal is to achieve the best risk-adjusted return, starting a company is crazy, but that's not my goal. Apparently my conclusion: if nothing is done to improve rocket technology, we will be stuck on Earth forever.
At the time, the big aerospace companies had no interest in radical innovation, and all they wanted to do was try to improve their old technology a little bit every year. And in fact, sometimes their technology gets even worse. Especially when it comes to rockets, the situation is rather bad.
In 69, we can also land on the moon with Saturn V. The Space Shuttle could then only take people to low-Earth orbit. Then, the shuttle was decommissioned. I mean, the trend is basically going to zero.
From 1969 to 1972, the American "Apollo Program" used the powerful Saturn V rocket to send a total of twelve astronauts to the lunar land in six times
If people think that technology will automatically advance every year, then this is not the case. It will only progress if smart people work like crazy and try to make it better. This is the real way for any technology to advance.
And as far as the technology itself is concerned, if people don't work hard, it will regress. You can look at the history of civilizations, numerous civilizations, such as ancient Egypt. They were able to build these amazing pyramids, and then they basically forgot how to build them, and then they forgot how to read hieroglyphs.
Look at the Romans, who were able to build these amazing roads, aqueducts and indoor pipes, and they later forgot how to build all of them.
History has many examples of this. I think you should always remember that entropy is not on your side.
Am:
What I like about you is that you are unusually fearless, and you are willing to do crazy things that others say. I know a lot of pretty crazy people, but you're still prominent.
Where does this come from? Or rather, how did you make the decision when everyone told you it was a crazy idea? Or, where did you get the inner strength to make such decisions?
Elon:
Well, first of all, I want to say that I actually think I feel pretty strongly about fear. I'm not without fear, my feelings are strong.
But sometimes, when something is important enough and you believe in it enough, you just let it go despite your fears.
People shouldn't think: Well, I'm scared of this thing, and I shouldn't be doing it. It is normal to feel fear, and if you do not feel fear, there must be some kind of psychological problem with you.
Am:
You feel the fear and let the importance of the thing drive you to do it.
Elon:
In fact, to some extent, fatalism helps. If you accept probability, then it reduces your fears.
When I created SpaceX, I thought the odds of success were less than 10%. And I accept this, in fact, I will most likely lose everything. But if we can move forward, maybe we can make some progress. Even if we go out of business, maybe other companies will take the baton and move on, which is still a good thing.
Yes, and so is Tesla. I certainly think that the odds of a car company succeeding are extremely low.
Am:
Currently, how high do you think the probability of establishing a colonization of Mars is?
Elon:
Well, oddly enough, I think the probability is quite high.
Am:
So when can I go to Mars?
Elon:
Well, for now, I'm sure there's a way, I'm sure success is possible for building a self-sufficient, growing colony of Mars, and I'm sure it's possible. Even a few years ago, I wasn't sure if there was a chance of success.
Send enough people to Mars, and I think it's possible to do it in about 10 years. Maybe faster, maybe 9 years.
Until then, I need to make sure that SpaceX doesn't go out of business and that I don't die, or that if I do, someone will take over and continue this.
Am:
You shouldn't follow the first launch to Mars.
Elon:
That's right. But anyway, the first launch is definitely not manned.
Am:
I want to go to Mars, but I can't stand the network delay.
Elon:
Yes, the network latency will be noticeable.
Mars is about 12 lightminutes from the Sun, while Earth is 8 lightminutes. The closest distance between Earth and Mars is 4 light minutes and the farthest is 20 light minutes. It's a little more, because you can't talk directly to the sun.
The relative relationship between the orbits of Earth and Mars
Am:
When it comes to the really important issues, artificial intelligence, you've been vocal about artificial intelligence. Can you talk about what a positive future for AI looks like and how we can get there?
Elon:
Okay, I want to emphasize that it's not that I'm advocating it, it's not a rule, it's just a prediction. Because sometimes people say, this is what I want to happen, not that this is what I think is probably the best of the existing alternatives.
The best alternative I can come up with, and perhaps the best anyone else can come up with, is that we democratize AI technology. That is, no single company or a small group of people can control advanced AI technology.
I think that's a very dangerous situation.
It can also be stolen by some bad guys, such as some evil dictator or state, who might send intelligence agencies to steal it and gain control. Once you get an extremely powerful AI technology, it becomes a very volatile situation, and you don't know who will control it.
I don't think the risk is that AI will develop its own will out of evil. The concern is that some people may use it for bad things, and even if they don't intend to use it for bad things, someone will take it away and use it for bad things. I think the danger is considerable.
So I think we have to democratize AI technology and make it widely available. And that's obviously why the whole team created OpenAI. That is, to help spread AI technology so that it is not concentrated in the hands of a few. But of course, this needs to be combined with a high-bandwidth interface that addresses the cerebral cortex.
ChatGPT lists Musk as a controversial figure = YES, requiring special treatment = YES. Lao Ma tweeted in response"!! ”
Am:
Humans are so slow.
Elon:
Humans are so slow, yes, that's right.
However, our brains already have the cerebral cortex and limbic system. The limbic system is like a primitive brain, it's like your instinct. Then, the cerebral cortex is the part of the brain responsible for thinking.
The two systems seem to work well together. Occasionally, your cerebral cortex and limbic system diverge, but in general, they both work well together. I haven't found anyone who wants to cut off their cerebral cortex or their limbic system.
Am:
That's right.
Elon:
Yes, this is very uncommon.
If we can improve the brain-computer interface between the cerebral cortex and the "digitally extended you" and effectively integrate with AI, as I said, the "digitally extended you" already exists, but there is a bandwidth problem, and then you are in fact a symbiote of artificial intelligence and humans.
If this solution is widely available and anyone can own it, then we have solved the problem of control. We don't have to worry about some evil dictatorial AI anymore, because we are AI as a whole.
This seems to be the best result I can think of.
Am:
You've seen a lot of other companies start small and get really successful.
Hopefully, I don't forget to ask this question in front of the camera, as a six-month-old company, how do you see OpenAI going?
Elon:
I think it's going pretty well.
I think we have a very talented team at OpenAI. Yes, a very, very talented team who work very hard.
OpenAI is structured as a 501c3 non-profit organization. Many nonprofits don't have a sense of urgency, and that's fine, they don't need to have a sense of urgency. But OpenAI has it because people truly believe in its mission. I think it's important that this mission is about minimizing the risk of future existential harm.
I think it's going pretty well, and I'm very impressed with what people are doing and the level of talent. Obviously, we're always looking for great people who believe in this mission.
Two co-founders of OpenAI, Sam Altman and Elon Musk
Am:
It was close to 40 people.
Okay, before we wrap up, a few more questions, how do you allocate your time? Where do you spend most of your time?
Elon:
My time is mostly spent at SpaceX and Tesla.
Of course, I try to spend some time each week on OpenAI. Basically, most weeks I spend half a day on OpenAI. Then, there will be some OpenAI transactions in the middle of the week. But beyond that, it's basically SpaceX and Tesla.
Am:
So, what do you do when you're at SpaceX or Tesla? Where do you spend most of your time?
Elon:
That's a good question.
I think a lot of people think I must have spent a lot of time in media or business matters. But actually, almost all of my time, 80% of my time, is spent in engineering and design, engineering and design. Developing the next generation of products took 80% of my time.
Am:
You may not remember this incident from a long time ago, but many years ago, you showed me around SpaceX. The most impressive thing is that you know every detail of rocketry and engineering. I don't think a lot of people can do that.
Elon:
Yes, I think, a lot of people think I'm a businessman or something, no problem.
But at SpaceX, Gwen Shotwell is COO, where she manages legal, finance, sales, and day-to-day commercial activities. Then I spent almost entirely my time with engineering teams, working on improving the Falcon 9 and Dragon spacecraft and developing a Mars colonization architecture.
Then, at Tesla, it was working on the Model 3. Sometimes in the design studio, usually half a day a week, deals with aesthetics, interface appearance. Then most of the week is spent following up on the engineering work of the car itself, as well as the engineering work of the plant.
My biggest feeling this year is that what really matters is the machine that produces the machine, which is the factory. And this is at least two orders of magnitude more difficult than the car itself.
The body shop of Tesla's Shanghai Gigafactory. Source: Tesla 2022Q4 financial report ppt
Am:
It's incredible to see robots work here and these cars are built like that.
Elon:
Yes, the degree of automation here is low compared to the upcoming Gigafactory and Model 3 production line.
Am:
How fast do these cars travel on the production line?
Elon:
In fact, the average speed of the production line is very slow. Including X and S, it may also be only 5 centimeters per second, which is very slow.
Am:
So how fast are you targeting?
Elon:
I believe we can reach at least 1 meter per second and speed up 20 times.
Am:
That's fast.
Elon:
Yes, that's less this speed.
I think it can be seen that 1 meter per second is slow walking, or moderate walking. Fast walking is 1.5 meters per second. The fastest humans can then run more than 10 meters per second.
If we only advance 0.05 meters per second, that's really slow. And at 1 meter per second, you can still walk faster than the production line.
(above)