 Our next speaker will be James Carroll. James has a PhD in computer science and a minor in ancient Near Eastern history. As a graduate student, he taught Pearl of Great Price. Isaiah in the Book of Mormon in the BYU Ancient Scripture Department. He is currently working at the Los Alamos National Laboratory doing ensemble machine learning research and computer-assisted radiographic analysis. That's a lot of words. For nuclear stockpile stewardship. Please welcome James. Today, I would like to talk about, while we're looking for my slides, the pace of technological change. This is a debate that has something of a long-standing history, especially in transhumanist circles. And I know that everyone here is probably familiar with the idea of the astounding nature of exponentials in the sense I'm speaking to the choir. But they really do do some pretty amazing and surprising things. For example, if I take a chessboard, put one grain of wheat on one square, then two on the next, four on the next, eight, 16, 32, et cetera, it sounds like you're not going to get a lot of wheat. But if you actually do this, this is the math. You add all those up. Those are some pretty big numbers at the bottom. And you end up with this 18 quintillion. That's actually enough wheat to bury India, the entire nation of India, about five feet of wheat. If you lay them end to end, you can go to Alpha Centauri and back four times. So that's pretty cool, right? Exponential curves do some pretty surprising things. And some people have noticed that certain technological events look like they're exponential. The most famous is Moore's Law. So the question is, how long is that going to continue? Well, so Ray Kurzweil has noticed this exponential trend and said some pretty amazing things, for example. This is the ENIAC, the world's first programmable computer. In about 50 years or so, we ended up with an iPhone, which is millions of times smaller, millions of times cheaper, and a billion times faster, and it fits in the palm of your hand. If that rate of change continues, in another about 50 or so years, it will fit inside a red blood cell. It will keep you alive. It will keep you healthy. So he claims and potentially grant you essential immortality. And that depends on the exponential change increasing. People who criticize him often try to talk about the complexity and difficulty of the problem. And those people simply do not understand the rate of exponential change. If exponential change is and remains exponential, he will be right. The real question is whether the rate of technological change really is exponential and whether it really will continue. Because if the rate of change is linear, we will not do this. We won't even come close. It will be thousands of years if it's linear, if at all. If it's exponential, it doesn't matter how hard the problem is. We will solve it very rapidly in a way that will surprise you. It could be a billion times harder, and we're talking maybe 30 more years. That's all a billion times harder buys you, is 30 more years. So let's forget how hard the problem is. It does not matter. And ask the question, is this really exponential or not? Because that's all we care about. That make sense? So here's another one of Kurzweil's amazing predictions. That in 2045, about when we have nanobots that may keep us alive forever, we also have computers for 1,000 bucks that are about as powerful as every human brain on the entire planet. Or that could simulate every brain on the entire planet. So is he right? Well, let's start with the arguments for linear growth. The first of them is that technological change happens according to this idea of low-hanging fruit. You pluck the low-hanging fruit first. And so the further you go, the harder it gets, and the slower it goes. It looks exponential at first, but then it slows down as you start trying to pick higher and higher fruit. That's the low-hanging fruit argument. And what's the evidence for that? We have some, actually. This is gene sequencing. It was beating Moore's law for a long time. People were really excited about the fact that it was beating Moore's law. And then it bottomed out. And it's hardly moving anymore. So there could be some evidence for this. The other comes from Wikipedia. This one's funny. This is the total active editors. This is the edits. And this is the growth in new articles. All of these things are showing a deceleration and change and movement. But perhaps the most interesting argument they make, and I think the most convincing, is this idea of life impact. And this idea says, even if technology continues to evolve exponentially, the impact on our lives is slowing. Even if I make new inventions at an exponentially increasing rate, it doesn't change my world as much as the inventions that came before. Frankly, what name and invention in the last 100 years that has changed people's lives more than indoor plumbing? And when you start thinking about cholera and the thousands of people who died because of lack of clean water, you realize there is none. We have not created an invention in the last 100 years that has had near this impact, even though we've created vastly more complex technologies, exponentially more difficult technologies like the internet that haven't had the same impact on our life. So here's another example. Let's talk about communication technology for a while. Before that, we couldn't even communicate. If someone moved across the nation, you probably never saw them again. If your daughter left because she married Joseph Smith and then wandered off to Nauvoo, you never talked to your parents again, right? And then we get the Pony Express, and some communication becomes possible, and that really changes people's lives because they can stay connected with people that they've otherwise would have lost. And then we get the Telegram, and that changes everything because now I can communicate with them the same day. That really has an impact on my life. And then we get the phone. They were really annoying. You had to share a line. The quality wasn't very good. You had to decide who wanted to talk to who by how many rings there were, longs and short, and there was no privacy. Everyone on the line could listen in. But you could talk to somebody. I mean, that was cool, but no real long distance. Well, then we improved. Now we have phone numbers and lines and long distance, and the phone itself improved. And then the phone design improved so that it was easier to use and easier to handle and easier to hold. And then we started getting buttons so your finger didn't hurt as bad. Dialing the little rotary thing. And then, you know, we get cell phones. Now you can take them with you, but they're big and they're heavy, and then they get small in light, and now you can play Angry Birds on them. And then it gets really cool because now I can see the person I'm talking to. Now think about what's happening here. The bandwidth requirement for each of these technologies is rising exponentially. To see someone's face is exponentially more bandwidth. So that's that exponential green curve. That's what's happening to the complexity of each technology as we move forward. But what's happening to the impact? Does it really change my world that I can see the person I'm talking to? It's useful. I mean, especially when I wanna show my dad something that my cute son is doing. But did it change the world the same as when I couldn't talk to my parents at all? Right, so the idea here is that the change and impact on our life is asymptoting out, not growing exponentially, even though we're creating exponentially harder technologies. Here's another example. This is a flight, right? So in 1903, the Wright brothers flew about seven miles an hour. About 50 years later, we were going about 570 miles an hour on a Boeing 707. The latest iteration of this is the Dreamliner. It turns out that supersonic speed has been a bust. It's not worth the cost. And it didn't make money. So now we're flying Boeing 787 Dreamliners 50 years later and they go about the same speed. And they're exponentially more complicated machines though. They really are. I mean, the composite materials and the complexity of the computers that drive them. This is an exponentially more complex technology. And it really doesn't get you there any faster. And the experience of going there is better. You get to watch a movie, you know? And it saves gas. It's much more fuel efficient. But has it really changed the world the same way as being able to get up, get on a plane and fly to the MTA conference even if you live in, where do you live? London, right? Europe somewhere, right? I was, that changed the world. The 707, it just made the world a little bit better. And yet it's exponentially more complex. So what are the arguments for exponential growth now? Now I've given the arguments for linear growth. What are the arguments for slowing growth? What are the arguments for exponential growth? The first is combinatorics. The idea here is that each technology is a combination of previously existing technology. So let's look at the stone ax. It's pretty much one thing, right? That was our starting technology. But then we built a hammer. And you could use a hammer to cut down trees to build saws. And you could use saws to cut down trees better to build better hammers, to build eventually log mills. And so the idea behind this is that we use each technology to build the next. And you look at the computer mouse and compare it to the ax. And it's made out of thousands of different parts, lots of different materials. It's a combination of the technology of digitization, communication, lasers, plastics. And you put all those together and you create a pointing device that's extraordinarily complex, exponentially more complex. And so the idea behind this is that the more technology you have, since new technologies are built out of combining previous technologies, the more low-hanging fruit you get. So this counters the low-hanging fruit argument because it says you get a larger amount of adjacent possible. In other words, as technology grows, think of the circle as the amount of technology we have. As it grows, the area around the edge of the circle grows and so there's more room for growth. And so it grows faster and faster and the low-hanging fruit argument is wrong. This is the combinatorics and adjacent possible argument. The next is based on history. This is Kurzweil's favorite. And what he tries to say is that if exponential trends have been going for 100 years, they're likely to go for the next 10. If they've been going for 1,000 years, they're likely to keep going for the next 100. If they've been going since the dawn of the universe they're likely to keep going for the next couple billion years at least. And so he tries to look for exponential trends that go back far in history and the further back they go then the further forward they go, right? So the first one is Moore's Law. Moore's Law is the most famous exponential trend and how does the laser work? Red button, ha. Okay, so here we go. Here's Moore's Law. It's an exponential trend. It's very famous but Kurzweil points out it's just one of a paradigm that went clear back to the 1900s. And so his argument there is that Moore's Law when it hits the quantum limits which it will about 2020, 2025, there will be something to replace it because there was something before it. That's his argument, right? So the further back the trend goes the more likely it is to continue. Here's another one. If we can go back to the dawn of the universe it gets even better, right? So this goes all the way back to the formation of the Milky Way Galaxy and all he's done is he's taken what he considers world changing events and plotted the time to the next world changing event and noticed that the time to the next world changing event is dropping exponentially. Therefore the time to the next world changing event will speed up and speed up until we have world changing events every second and then singularity. He's done that with several different combinations of world changing events. Here's a different one, same trend. Then you took a bunch from a bunch of different people put them all together, same trend. So who's right here? This is a debate that as transhumanists will vastly impact what we predict about the future and whether what we hope for in terms of the creation of God or becoming like God if you already exist depending on which view you take from the earlier talk how that works will depend and how it will unfold will depend on what view you take with regard to this argument. It is the most important argument in my view for determining what the future is gonna be like. Well let's start with this one. Life impact. They're right that the linearists are right. We haven't created something nearly as impactful as indoor plumbing in a good 100 years. However, there are two technologies on the horizon. The first being artificial intelligence and the next being biotech that I believe have the potential to be as earth shattering as indoor plumbing. And so if you believe that maybe things kind of start and bump and slow down we may be next to the next big bump that will change everything. And so again, there's some mixed messages here. Yes, impact is slowed. No, that doesn't mean it's gonna stay slow. We may be on the verge of something really big here. We should pay attention. So I'm gonna keep score here. And I'm gonna say that the exponentialists won that because there is a big impact technology potentially right on the threshold of being able to do. But let's talk about this one. The easiest counter to this argument is that Kurzweil suffers from something I'm gonna call presentism. Presentism is the idea that when I start listing things that were important to me, the further back you go, you get an exponentially less smaller chance that that thing will be important to me. And so what Kurzweil might be noticing is just exponentially less important events to him happened at the Big Bang. And to everyone who makes this list, that's true. It may have nothing whatsoever to do with the rate of technological change. It might, but it might not. And so trying to sort that out, it turns out, is really difficult. How do you divulge yourself of all presentism bias and really figure out if things are improving exponentially and have been for eons? Well, so one of the ways to do this is to limit yourself to phase shifts, what I'm gonna call a phase shift. It's an exponential, it's a point where you take the level below you and use it to create a fundamentally new level that then gets combined to create another fundamentally new level. And if we limit ourselves to those events in history, maybe we can avoid presentism. So let me show you what I mean. If you saw my talk last year about the existence of God, you know what I mean about specialization, cooperation and combining levels to create the next level. So we start with the Big Bang. That's the most obvious beginning point. And then at some point in history, the quarks and gluons all kind of cool and they start forming atoms. So from a lower level, we get a new level with new features and complexities we're gonna call the atom. Then atoms combine to form molecules. There's two ways to do this. At the moment they first form molecules at the moment they first form self replicating molecules. I'm gonna do the replicating version. The reason I picked that is because I want to know when that level will create the next level down the replicating chain because that's the chain that leads to us and it creates our progress. So we get atoms to self replicating molecules, self replicating molecules to single cells. Now part of the problem is we don't know when the self replicating molecules happened relative to the cells. So we'll have to guess about some of these dates. Then we get the single cell. Then something really funny happens. A bunch of single celled organisms merge together. They specialize and they cooperate and they create something new. A new single organism that is actually a collection of the previous level of organisms and that's the eukaryotic cell which has cells essentially inside of cells. And then we get the multicellular organism where many eukaryotic cells merge together to create a brand new multicellular organism and each of them begin to specialize and trade just like our skin specializes in one thing and our teeth specialize in another, right? And then there's another level where we as a group of multicellular organisms merge together and create some sort of collective intelligence. And that's something that hasn't quite happened yet. It's in process. So picking this date is really hard because you could pick when we first create hunter gatherer societies, you could pick when we first create city states, you could decide when we first create the printing press, when we first create the internet. You could pick a lot of dates for that because it's a smooth process that is still in process and will continue I believe as we become ever more tightly integrated into a collective organization. So I chose the internet for that one. It's just a random guess. And now the problem comes when is the next thing going to happen? Before I can even put this one on the curve, I need time to next. I mean, that's how they're plotted, right? And we don't even know what the next level is. So I can't put it on the graph. So I'm gonna use this one to put this one on the graph in terms of time to next and then plot these over here. And when you do that, this is what you get. And it's not exponential. It has no clear pattern even as far as I can tell. And so this may not be the answer some of you want but the only technique I know to throw away presentism leads to a conclusion that I have no idea if there really is an exponentially increasing change that goes all the way back to the dawn of the Big Bang like Kurzweil would really like. It just, it may be there, it may not be but I can't tell. And I think that there are people who really have an opinion about this and I always kind of smile at them whatever your opinion is. And I think you think you know more than you know if you take a side stand on this because it's just not nearly as clear as Kurzweil would like. So I'm gonna add one to the uncertain category here. This is one the linearists use because they like to show that the exponential progress of gene sequencing has stalled. That's important because biotech was one of the two that I said was on the horizon that could really change the world, right? Well, there's a caveat hidden in this graph and it's right here in this word at the top price. It's not the cost, it's the price. So we've begun to specialize to the point where if you want sequencing done, you tend to take your gene to a bank and they sequence it for you because they can do it better than you can even though you can buy gene sequencing kits. It's cheaper to let them do it in mass production now. That was an interesting shift and guess when it happened, right, about here. And what that means is these people are using the economics of scale to really sequence fast and to make it cheap but price is not just a function of the cost of doing something, it's a function of price, cost and demand, supply and demand. And what happened right about here was an exponential rise in demand for gene sequencing and so the price stalled unsurprisingly. We are now meeting an exponentially rising demand with a flat cost and that's pretty impressive. So I think this argument actually fails the linearists, I don't think it shows quite what they think it shows. It's hard to tell because these in-house groups tend to keep their costs secret so we don't actually know, I believe, what's happening under the hood here anymore. And the result of that is another, either a failure for the linearists or an uncertainty, it's hard to tell. But I'm gonna put that one in the uncertain category. Now it's at least a failure for the linearists but it may not be a score for the exponentialists that we don't know yet. The next one is this pretty graph of Kurzweil's. It's really fun because he claims there's not just an exponential but he claims there's an exponentially rising speed of exponentiating. He's got a double exponential. The exponential, it would be linear on this log plot but he's got an exponential curve on a log plot which is an exponential, exponential. And that really changes things, right? However, I've looked at this data, the raw data for this and I've gathered data since, this was done in 2005. Data that kind of moves off into this direction and I've tried to ask the question, is he right about the double exponential? And what I believe I see are two exponential trends with a phase shift. There's an exponential shift right here or trend right here and then there's a new exponential trend that takes off right here and the newest data seems to follow this red line not up Kurzweil's double exponential. And if you start asking questions, why would that be, do you think? If you look at the date, the date here is the invention of the ENIAC. And so right at the invention of the ENIAC, there appears to be, the first programmable computer, a phase shift in the rate of exponentiation and it's been strictly single exponential sense. So to use this double exponential to extrapolate I think is a mistake. There's something else going on here and that's about 2025, Moore's law will hit its end. Kurzweil claims that it's just one trend, we'll find another trend, that may be, but there is something I think fundamental about making things smaller. When you hit the limits where you can't make it smaller no matter what new technology you move to, something has got to change either in the way you build the computer or any number of things. So at Moore's law, I think we get another phase shift and it'll be hard to say what happens. So I'm gonna call that one uncertain because if you come up with the right technology you could really take off again or it could flat line for all we know. So we've got something else we don't know about. All right, I need to go faster. Microprocessor clock speed, Kurzweil predicted it was gonna go off exponentially, that was his curve. He wrote his book in 2005. These numbers are predictions based on a white paper he found, it turns out we started melting our chips if you raised the clock speed and because of heat issues, clock speeds bottomed out right when he wrote his book he was wrong. So this is a win for the linearists. However, our fastest supercomputers just kept on going. How did we do that? Well, we started adding more, our dyes became more efficient even though they didn't have faster clock speeds and we added more chips per die and the supercomputers added more machines to the network. And so this is our number of nodes, number of cores on the supercomputers and that's took off exponentially and that's how we did that. And that means something very important, I think. We'll get to that in a second. Here's GPU supercomputers. These are also very important because they're parallel, they're a lot more like the brain and these could provide a technique for exponentiating beyond Moore's law. So we've got price and speed, those also seem to be going exponential in Shono times of stopping. So here's the money graph. This is the amount of computing approximately it takes to simulate a full human brain and you'll notice where they cross. This graph has continued exponential despite the end of clock speeds, despite everything else and it crosses before Moore's law quits. So even if Moore's law quits on us, it's gonna quit too slow to give us the hardware we need to solve the problems we wanna solve. In other words, that big next technology I told you about, we will have the hardware to do it before Moore's law quits. So no matter what happens after Moore's law quits, it's too late to stop this, this AI technology. Now we have to deal with software and that's a subject for another talk. So today was about hardware. The software to do this is much more tricky to predict and we'll try to do that maybe next year. I'm not sure what we'll do but we're gonna quit here. Thanks.