AI Will Create New Wealth, But Not Where You Think | Carnegie Mellon University, Po-Shen Loh (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) The AI today can do those problems like (00:00:03) this. Actually, at this point, even a (00:00:06) very sophisticated math coach can be (00:00:08) replaced by the AI tool if you decide (00:00:11) you want to do it. As I started running (00:00:13) around rural areas, I walked into the (00:00:15) elementary school. I walked into the (00:00:16) fourth grade classroom and I was just (00:00:18) going to do my usual thing, Mr. Poe, the (00:00:20) substitute teacher. So, I wrote on the (00:00:22) blackboard, what is 1 + 3 + 5 + 7 + 9 (00:00:29) equals? As soon as I wrote equals, (00:00:32) behind me, I heard a bunch of kids (00:00:33) yelling 25. I've actually never (00:00:36) experienced before, a classroom. This (00:00:39) was one of the best classrooms I have (00:00:41) ever taught. And I just told you the (00:00:44) profile of the area was high poverty. I (00:00:47) asked the person who took me around (00:00:49) afterwards, "These kids are amazing. Do (00:00:52) they play games on their phones?" And (00:00:53) the lady told me, "They don't have (00:00:56) phones. It's because of the money. In (00:00:58) fact, they might not even have the (00:00:59) internet access." Then I said, "What do (00:01:00) they do for fun?" And she said, "Well, (00:01:03) they just figure out how to make their (00:01:04) own games." The problem was these people (00:01:07) who I met, people outside in the rest of (00:01:09) the world, didn't know about these (00:01:11) particular great people. So then I (00:01:14) realized actually that huge pool of (00:01:16) authentically interested and curious (00:01:18) kids throughout rural America, this (00:01:21) could be an enormous untapped potential (00:01:23) across not only the US but the the (00:01:26) entire world. This could potentially (00:01:27) create a totally new economic flow (00:01:29) system. This might just be what we need (00:01:33) for the 21st century after AI. I will (00:01:36) also say for everyone who wanted a (00:01:37) stable life, good luck cuz AI is going (00:01:40) to take that. (00:01:42) Hi, I'm Po Shan Lo. I'm a mathematician (00:01:45) who has gotten very distracted by the (00:01:47) real world. And now my main focus is on (00:01:50) trying to build a more thoughtful world (00:01:52) to help humanity survive after AI. (00:02:09) These days, my biggest focus is on how (00:02:12) to bring opportunities to places where (00:02:15) there might not have been as many (00:02:16) opportunities before. The latest (00:02:18) direction that I've been pushing in has (00:02:20) actually been rural communities (00:02:22) throughout the United States. I went to (00:02:24) South Carolina last year in December. I (00:02:27) went into a classroom of fourth graders. (00:02:29) As we were driving down the road, it was (00:02:31) very clear from the stores and the (00:02:33) buildings that this was an area that was (00:02:36) quite impoverished. Well, I walked into (00:02:38) the fourth grade classroom and I was (00:02:40) just going to do my usual thing, Mr. (00:02:42) Poe, the substitute teacher. So, I wrote (00:02:43) on the blackboard, "What is 1 + 3 + 5 + (00:02:49) 7 + 9 =?" As soon as I wrote equals, (00:02:53) behind me, I heard a bunch of kids (00:02:55) yelling 25. And we talked for about 20 (00:02:58) minutes. And through the whole thing, (00:02:59) the kids were all suggesting ideas. They (00:03:02) were also very respectful of each (00:03:03) other's ideas. If someone was giving an (00:03:04) answer, I said, "Oh, let's let's all (00:03:06) listen to that person." Everyone did. It (00:03:08) was a marvelous class. It was one of the (00:03:10) best fourth grade classes I've ever (00:03:12) taught. And I just told you the profile (00:03:14) of the area was high poverty. Uh I'll (00:03:17) also say the ethnic makeup of the (00:03:18) classroom. It was 95% African-Amean. I (00:03:21) asked the person who took me around (00:03:23) afterwards, "These kids are amazing. do (00:03:26) they play games on their phones? And the (00:03:28) lady told me they don't have phones. (00:03:31) It's because of the money. They don't (00:03:33) have the phones. In fact, they might not (00:03:34) even have the internet access. Then I (00:03:36) said, "What do they do for fun?" And she (00:03:38) said, "Well, they just figure out how to (00:03:40) make their own games." And I found out (00:03:42) that throughout all of rural America, (00:03:44) which is where I've been running, wow, (00:03:46) there are so many kids who are actually (00:03:49) really, really interested in challenging (00:03:51) themselves. It's just that because of (00:03:53) the way the curriculum is designed, they (00:03:55) haven't been actually standard (00:03:56) curriculum is designed just to make sure (00:03:59) that you know how to do a standard (00:04:00) problem. In this future world, we need (00:04:03) people who can do non-standard problems. (00:04:05) Well, it turns out that there's plenty (00:04:07) of kids who were really poised to do (00:04:08) that. Actually, it even felt more (00:04:11) authentic than what I found in cities. (00:04:14) In December, I went to Africa because I (00:04:17) was particularly interested in Africa (00:04:18) because the population is getting bigger (00:04:20) and bigger and as a fraction of the (00:04:22) world's population is going to become (00:04:24) more and more significant. When I went (00:04:25) there, I wasn't going to propose (00:04:28) particular solutions. I was going there (00:04:30) to learn and try to see what was going (00:04:31) on. And immediately I saw lots of very (00:04:33) capable people. Then I started to think (00:04:36) why is the economic development not as (00:04:38) strong given that there are all these (00:04:40) great people. The problem was these (00:04:42) people who I met people outside in the (00:04:44) rest of the world didn't know about (00:04:46) these particular great people from their (00:04:49) outside world. It was just I guess there (00:04:51) is a place called Africa. How do you (00:04:53) send resources? Okay, if you send it (00:04:54) this way, somebody is professional at (00:04:56) receiving resources. That's not as good (00:04:58) as knowing this person can really use (00:05:00) the resources. So with the whole system (00:05:03) we do, we have high schoolers who coach (00:05:05) middle schoolers. The high schoolers can (00:05:08) be from anywhere in the world and they (00:05:10) are selected through our our method (00:05:12) where we try to find who really cares (00:05:13) about people and who's also very good at (00:05:15) thinking about ideas on the spot to (00:05:17) solve questions. They always teach in (00:05:19) Paris. By having the two of them teach (00:05:21) in Paris, they get to know each other. (00:05:23) If anyone in say United States or Korea (00:05:26) or or China or Canada or Europe, if they (00:05:30) partner with one of these people from (00:05:33) Rwanda or Ethiopia, any other country (00:05:35) which we might start to have a (00:05:36) relationship with, well then they would (00:05:39) very quickly realize, oh, each of us are (00:05:42) good. My prediction is that 5 to 10 (00:05:45) years down the line, they'll be looking (00:05:46) for jobs. It turns out that you can have (00:05:48) remote work. Remote work does work well. (00:05:50) The only issue is who would you hire? (00:05:52) Hm. If the person who was living in the (00:05:54) in the developed country was starting a (00:05:57) company or if they were looking for (00:05:59) partners or people to work or even (00:06:01) looking for employees to help work, I (00:06:03) anticipate that they might call up the (00:06:06) person that they know. There's even an (00:06:08) economic arbitrage. The amount of money (00:06:11) that people are used to earning in (00:06:12) different countries is very different. (00:06:14) In fact, if you split the difference, (00:06:16) both sides win massively. one side will (00:06:19) save a lot, the other side will get to (00:06:20) live really, really, really well. So (00:06:23) there was just the observation I made of (00:06:25) h given that there's this huge economic (00:06:28) difference in how much $1 can buy in (00:06:31) each country, if people who are really, (00:06:33) really, really good are detected by (00:06:35) other people. Well, then they could get (00:06:38) they could get a really good remote work (00:06:40) job, then they might be in a situation (00:06:42) where they can start their own thing. (00:06:44) Sometimes to start your own company or (00:06:46) start your own initiative, you need to (00:06:48) first be safe. You might also have a (00:06:50) whole bunch of friends you know from (00:06:51) doing all of this co-eing at that stage (00:06:54) those friends in the United States or in (00:06:55) the rest of the world. The only (00:06:57) difficulty is can it get through a (00:06:59) network of trust and if there's a direct (00:07:01) trust where people know that person (00:07:03) really can do stuff well then that (00:07:05) person could start to get resources (00:07:07) coming in through this network from (00:07:09) developed countries to developing (00:07:11) countries which are shortcutting. (00:07:13) they're going directly to people who can (00:07:15) use it. So this could potentially create (00:07:17) a totally new economic flow system. This (00:07:20) could also work in developed countries. (00:07:22) In general, to me, this is the value of (00:07:24) a huge network. I'm a network theorist. (00:07:27) So when I thought of this whole thing, I (00:07:28) said ultimately building a high trust (00:07:31) network among lots of people who love (00:07:34) helping other people and love thinking (00:07:37) hard. This might just be what we need (00:07:40) for the 21st century after AI. (00:07:48) I think that what we're observing is (00:07:50) whatever people are good at all these (00:07:52) skills, the AIS are getting better at (00:07:54) them and eventually getting better than (00:07:56) them as well. For a while, people were (00:08:00) saying that the safe job will be plumber (00:08:02) like a blue collar job. But if you look (00:08:04) at how many humanoid robots there are, (00:08:07) there are a lot of them. Oh, actually (00:08:09) one of the very famous US companies was (00:08:11) Boston Dynamics. As soon as I saw that (00:08:13) Hyundai had bought them, I know what (00:08:15) Hyundai wants to use those robots for. (00:08:17) Not for dancing. Hyundai manufactures (00:08:19) large amounts of stuff. Hyundai probably (00:08:22) would be very happy to be able to have (00:08:23) tons of robot workers. Hundai also has (00:08:26) money. So, it's not going to be very (00:08:27) long before you got these humanoid (00:08:29) robots that can make all kinds of stuff (00:08:30) that are working in all the Hundai (00:08:31) plants. that's going to that's going to (00:08:33) wreak havoc across the blue collar as (00:08:35) well. So then I started to think what is (00:08:38) special about people. One of the things (00:08:39) I think is quite special about people is (00:08:41) that hopefully they care that humanity (00:08:44) still exists. And the best part is the (00:08:46) ones who do if you talk to them you can (00:08:48) read it from their eyes. Beautiful thing (00:08:50) about humans is that you can tell when (00:08:52) you talk to someone this person cares (00:08:56) about the big picture more than just (00:08:59) about themsel. You will never be able to (00:09:01) get that confidence looking at a robot's (00:09:03) eyes. I see a lot of electric vehicles (00:09:05) on the road today. A lot of EVs. An EV (00:09:09) is basically a computer with four (00:09:10) wheels. Why I say this is because I'm (00:09:12) emphasizing one of the most important (00:09:14) parts of the electric vehicle is the (00:09:16) computer. Many of the electric vehicles (00:09:18) get constant software updates. What (00:09:20) would happen if somebody hacked into the (00:09:23) software update system? Next week, one (00:09:26) particular brand of EVs at 5:30 p.m. (00:09:29) they all accelerate to full 100%. (00:09:32) The more interconnected our world is, (00:09:34) the easier it is for one move to cause (00:09:37) very bad things to happen. And if you (00:09:40) ever tried editing code, you know that (00:09:42) it's actually possible to make weird (00:09:45) things happen without even fully (00:09:47) understanding, especially if the code (00:09:49) was written with AI. So the car which (00:09:51) was supposed to help you can change into (00:09:53) the car that was supposed to hurt you. (00:09:55) You have absolutely no way of knowing (00:09:57) because it has no eyes. It has has no (00:10:00) eyes you can see that don't change. That (00:10:03) will also give a job opportunity because (00:10:05) there will need to be people who you can (00:10:07) trust to take care of things and to make (00:10:09) sure things are safe. You want to know (00:10:11) that the people you put into these (00:10:12) positions care about things that are (00:10:14) bigger than themselves and they aren't (00:10:15) easily bought off by someone bribing (00:10:17) them for a million dollars or something (00:10:18) like that. So my prediction is that the (00:10:21) kinds of people that are going to have a (00:10:23) lot of job opportunities are the ones (00:10:25) where you can tell that you can really (00:10:27) really count on this person and that (00:10:30) person is very flexible. Generally (00:10:32) speaking, when I hire people, if I meet (00:10:34) someone like that, I just try to think, (00:10:35) can I just find a place for you in my (00:10:37) organization? Cuz this kind of person, (00:10:39) you can plug into anything. Great (00:10:41) intention and great learning capacity. (00:10:44) They're going to work hard towards a (00:10:45) goal that's meaningful. Okay, let's (00:10:47) figure it out. I don't want to hire (00:10:49) someone who has been trained to do one (00:10:50) particular task because now I've (00:10:52) discovered wait one or two more years I (00:10:54) can use the AI to do that task and it'll (00:10:56) be way cheaper. People who for whom you (00:10:58) can see that they just want to do good (00:11:00) stuff. (00:11:02) We don't have enough of them. The more (00:11:03) automation there is the more things that (00:11:05) can go wrong. We don't even have enough (00:11:07) good people to watch out for all this (00:11:09) stuff. I think that's one of the major (00:11:10) future needs for humans. (00:11:16) AI definitely creates a way for more (00:11:19) people to access education. Actually, I (00:11:22) get to learn all kinds of things right (00:11:23) now. My chat GPT history currently has (00:11:26) questions about what's in the Quran. (00:11:28) Okay, so the best thing is today if you (00:11:31) want to learn something, oh, you really (00:11:33) can learn it. You can you can ask these (00:11:35) AIs. In fact, this morning I came in (00:11:37) here on an overnight bus and I was busy (00:11:39) uh updating a particular website. is a (00:11:42) is an online video game. We're about to (00:11:44) put out video game meaning it's a math (00:11:45) game. It's for people to do math. But uh (00:11:47) there are all these math problems. In (00:11:49) the old days, I had to actually do the (00:11:51) math problems myself. I have all these (00:11:53) old math problems from other math (00:11:54) contests. The AI today can do those (00:11:57) problems like this. I was using Claude's (00:11:59) Opus 4.5, which is the very advanced (00:12:01) version. And I was having it solve all (00:12:03) these advanced math problems and give (00:12:05) hints. So I'm explaining like actually (00:12:07) at this point even a very sophisticated (00:12:10) math coach can be replaced by the AI (00:12:13) tool if you decide you want to do it. So (00:12:16) the big question becomes for the student (00:12:18) do I want to do this right because there (00:12:21) are any all you need actually is just to (00:12:23) open up your favorite AI and start (00:12:25) asking it all kinds of questions. I was (00:12:27) actually just in China uh last week and (00:12:29) I was at one school and that one school (00:12:31) was showing me how they had some AI (00:12:34) powered app. So that app was designed to (00:12:37) let you do the kinds of problems that (00:12:38) will appear on the exams so that you (00:12:40) could rank higher so to speak. So it was (00:12:42) it was an AI powered performance (00:12:44) improver for standard exams in China. (00:12:48) One of the curriculum people asked me, (00:12:49) "What do you think?" And I said, 'You (00:12:51) know, actually, if I was using AI to do (00:12:53) education, (00:12:55) I don't think I would do it that way (00:12:57) because I think that that's just (00:13:00) creating people who are (00:13:03) human versions of AI. You're just making (00:13:06) human robots. The most important thing (00:13:08) today is that you want to learn (00:13:11) something and then you're curious enough (00:13:12) to go and engage with it. Ah, but you (00:13:15) also need to be able to think. So that's (00:13:17) the other piece. It's very dangerous (00:13:19) today for people just to ask AI stuff (00:13:21) because the AI can tell you something (00:13:23) and it sounds authoritative but it could (00:13:26) be bogus. So the bottom line is that the (00:13:28) playing field for learning stuff at this (00:13:30) point if you just want to go and (00:13:31) interact with AI you can everyone can (00:13:34) have it. Then the deep question becomes (00:13:37) why would anyone do it? So that's why (00:13:39) the way I work now is on the philosophy. (00:13:41) In the old days, people would do it (00:13:43) because then you can get a higher rank (00:13:45) and then you can get into a better (00:13:47) university. But today, even if you do (00:13:49) that, you still can't get a job. It's (00:13:51) actually quite sad. A lot of people who (00:13:53) are running along this pathway, they're (00:13:55) going to work very hard for about 20 (00:13:58) years of their life being told by their (00:14:00) parents, you do this so that you have a (00:14:02) better job at the end. Finally, they (00:14:04) graduate and they still have no job. (00:14:06) That's going to be a that's going to be (00:14:08) a major mental health crisis. So instead (00:14:10) the the way I think is more healthy is (00:14:13) if people have a real intention to do (00:14:16) something that is bigger than themsel (00:14:18) that involves other people. That's why (00:14:20) what I love is when I see a kid whose (00:14:22) eyes are saying I want to help you. It's (00:14:26) so interesting. You can see this from (00:14:28) human eyes. They're going to be very (00:14:30) curious. They're going to keep learning (00:14:31) stuff. And the interesting thing is then (00:14:33) they can they can become arbitrarily (00:14:35) good. So when I was thinking about how (00:14:37) do you motivate people to want to be (00:14:39) thoughtful, it occurred to me thoughtful (00:14:41) people like being around other (00:14:44) thoughtful people. It's fun. If you if (00:14:46) you're a person who likes to help other (00:14:47) people and thinks about stuff and likes (00:14:49) to think about stuff, the moment you (00:14:51) meet anyone else with these two (00:14:52) characteristics, you very quickly become (00:14:54) friends and you become trusted friends. (00:14:56) Well, then these thoughtful networks (00:14:58) become very, very strong. So then I (00:15:00) realized there's another way to motivate (00:15:02) people. Find thoughtful people, connect (00:15:05) them to each other. Then they naturally (00:15:07) will start to try to find ways to create (00:15:09) value. Some of these people will also be (00:15:11) entrepreneurs. The heart of (00:15:13) entrepreneurship is finding pain points (00:15:16) in other people and solving them. And by (00:15:19) the way, you'll get money out of that (00:15:20) because you have solved a problem. (00:15:22) Right? That's how I realized networking (00:15:24) together thoughtful people provides a (00:15:26) 21st century way to provide ongoing (00:15:29) opportunity which is actually even (00:15:31) better than the central authority way of (00:15:35) everyone take a test go and see who (00:15:37) ranks the highest the ones who rank the (00:15:38) highest give them something that they (00:15:40) all can do independently right there (00:15:42) there's different ways that you can run (00:15:43) a society (00:15:45) different countries do have different (00:15:47) systems though so I I I actually see (00:15:49) this in some of the other countries that (00:15:50) I go to where due to the way the system (00:15:53) is everyone is just struggling along (00:15:56) competing to get these high scores on (00:15:59) exams because that's what's going to (00:16:00) determine uh their future. Then of (00:16:02) course the question is how can we help (00:16:04) them very strong practical advice would (00:16:07) be to learn English learn English to a (00:16:08) very high level of fluency because that (00:16:11) gives you access to almost well not (00:16:13) almost but it gives you access to a huge (00:16:15) world of opportunities. If the only (00:16:17) language you know is your native (00:16:18) language, you only get opportunities in (00:16:21) systems which use your native language. (00:16:23) And if the system is designed in a way (00:16:25) where most of the people will not have (00:16:26) opportunity, it could be beneficial to (00:16:28) be able to play in all the other (00:16:30) systems. Next thing is to pick up this (00:16:34) this this thoughtful thing where when (00:16:35) anyone meets you, they see, oh, you (00:16:37) really do care about other people. That (00:16:39) is actually what's going to make people (00:16:40) want to pull you out of those systems. (00:16:42) Because for example, if anyone has (00:16:44) fluent English and if you live in (00:16:45) another country, Korea for example, oh, (00:16:48) we'd love to have you in our program. (00:16:50) We're not designed to only take the top (00:16:52) 10%. We're designed to take all the (00:16:55) great people. We're great just means (00:16:58) that you actually like other people and (00:17:01) you actually want to do things for other (00:17:02) people. And if we give you weird math (00:17:05) problems that you've never seen before, (00:17:07) you can think about them. It's very (00:17:09) useful for people to take a step back (00:17:10) and think why is the system built this (00:17:13) way anyway? What was the intention? And (00:17:15) then you might need to break out of the (00:17:17) system. You might need to find your own (00:17:18) other way to do things. That's (00:17:19) entrepreneurship. I will also say for (00:17:21) everyone who wanted a stable life, good (00:17:23) luck cuz AI is going to take that. So (00:17:26) unfortunately anyway people need to move (00:17:28) to this direction. (00:17:31) I'm Mih. The title of the class is the (00:17:33) modern software developer. It's (00:17:34) definitely the first class in where the (00:17:37) focus is AI across the SDLC. It's the (00:17:39) first of its kind at Stanford within (00:17:41) like a few hours of the class being (00:17:42) announced and it kind of opened up for (00:17:44) enrollment filled up over 100 students. (00:17:46) Something kind of crazy is happening. (00:17:48) Software development and AI is really (00:17:50) starting to make its way into every (00:17:52) single part of of how software is being (00:17:54) done and and clearly something was (00:17:55) changing. What is interesting is that (00:17:57) there is this emergence of kind of like (00:17:58) a new I would say class of like engineer (00:18:01) which is like the AI native engineer and (00:18:03) AI is that language. AI is that new (00:18:05) language. This particular generation of (00:18:08) junior developers of junior engineers (00:18:09) the people that are now entering the (00:18:11) workforce will I think be the first kind (00:18:13) of generation of that new shift. A (00:18:15) single developer become a manager of (00:18:18) agents. So really knowing how to like (00:18:19) properly handle multiple agents is like (00:18:21) the last boss in the game. Like if you (00:18:23) can do that really really well then you (00:18:25) are like literally like the top top.1% (00:18:27) of of users even today.