This content was created by Tim Draper. This content was uploaded by Tim Draper. This is a YouTube talk titled "How to make sense of exponential change | Pascal Finette" (created 2016-11-15). It is a single-speaker presentation by Pascal Finette that explains technological exponential growth using Kurzweil's law of accelerating returns and S-curve dynamics, references Clayton Christensen's Innovator's Dilemma and Peter Diamandis' 6 Ds of disruption, and gives concrete examples (digital cameras, Uber). The talk also includes extrapolations about compute power reaching human-brain parity around 2030 and aggregated compute capacity by mid-century (~2050–2060), and discusses dematerialization, demonetization, and democratization of technologies.
Chapters
So let's talk a little bit about like the how. How do you make sense of this change? So you get an idea of like technology moves on an exponential curve. That's fine. We're seeing it happens in many other, many areas. Let's talk a little bit about the how. And let me start out with a disclaimer. And this is what Ray Kurzweil, one of our founders, calls the law of accelerating returns. The disclaimer is exponential curves are actually not really exponential into infinity. Take Moore's Law. Number of transistors per square inch per integrated— on an integrated circuit, right? If you take that to its logical conclusion, the transistor needs to be subatomic scale. Right? Like physics, at a certain point, physics don't work anymore. So what Ray did was, he was interested in this idea and said, let's look at actual data points. So 110 years of actual data, and he looked at— What's Moore's Law? He had a different metric. He looked at calculations per second per $1,000. How many calculations can I perform per 1 second per $1,000, which is kind of like the economic underpinning. And then this is a logarithmic plot, which is the reason why you see it as a linear graph. As you can see, Moore's Law is actually stable over a much longer period of time than what Gordon Moore predicted. It's stable over 110 years. So it goes from technology to technology to technology. Yeah. And what the underlying insight is, is that these technologies are S-curves, right? So they move on an exponential, but then they flatten out because physics don't work anymore. But what happens is that these S-curves stack on top of each other. So as one technology gets to maturity, the next comes in, the next comes in, the next comes in. And then as a stacking, they stay in a perfect exponential. Yeah. If you're familiar with Clayton Christensen wrote a book called The Innovator's Dilemma where he talks about the same thing, different perspective. Clayton's insight was to say a company which dominates these eras is never the company which dominates the next, or very rarely. Let's give you an example. Mainframe computers was IBM and GE and Honeywell. Minicomputers, Digital Equipment and Data General, like 2 companies which don't exist anymore. PCs were dominated by mostly Microsoft and the OEM manufacturers. The next one is handhelds and phones, which is obviously Apple and Google. The next era there is, you know, IoT, Internet of Things, many like very small devices. I can make a bet with you that I'm pretty sure that neither Apple nor Google will be dominant player in that space. Because it requires a very different thinking. The most exciting times is always when you go from one era to the next. Because that is where like, you know, billionaires are made. This is where fortunes are made. This is where you really drive change in the world. One fun thing you can do with this graph is you can extrapolate it. So this graph has been true for the last 110 years. Let's assume it stays true. There's not a lot which shows that it shouldn't. You can detect the moment, you can predict the moment when you have the raw compute power of a human brain, which happens in 15 years. So by 2030-ish, you will have a computer which has the raw compute power of a human brain. More fun, you can actually determine the moment when we have Skynet, a computer with the raw compute power of every single human brain in one machine, which happens around 2050, 2060. So we are moving super quickly into a really, really weird world where the compute engines around us will be smarter in terms of computing power than us. Let's look at this from a slightly different angle. Peter Diamandis, our other co-founder, has this model which is kind of like an explanation model. It's called the 6 Ds of disruption. The way to explain this is let's talk about the digital camera. This is the very first digital camera. This guy is called Steven Sasson, who's the guy who invented this. This camera had 100 by 100 pixels. 0.01 megapixels. Obviously horrible. Terrible. And also like not quite like, you know, your point-and-shoot format. What I love most is you save the images on a cassette tape. It's the precursor to the SD card. So here's how this works. You start out with— you take something which is analog, you digitize it. So what Steven did was he took film, which is a chemical process, and turned it into a digital good. In the beginning, it's deceptive. So first digital camera had 0.01 megapixels. What Steven actually managed to do is he moved it immediately on an exponential curve. He doubled the pixel count literally every year. So it went from 0.01 to 0.02 to 0.04 to 0.08 to 0.16. Now you can see even 0.16, 4 years later, is terrible. It's horrible. It's deceptive. Like the people who worked with Steven said like, this is not going anywhere. Because all they could see is the zero in the beginning. But then eventually it becomes disruptive. So now the moment when you hit 2 megapixels, you're getting disruptive. 2 megapixels is good enough for a 9 by 13 print or 6 by whatever it is in the US. This is the moment when I bought my first real digital camera, Canon Digital ELPH. And I never bought film again. Wow. Because it's good enough. Then you start to dematerialize. So what was a $500 camera now becomes a $1 piece of equipment which is bolted onto your phone, and everything else becomes software. Once that happens, you demonetize it massively. So again, you go from a $500 camera plus all the supplies, film and processing, to a $500 camera To— well, actually, they get it more or less for free. And then you democratize it. Now everybody has a camera. There's now 3 billion pictures uploaded to photo sharing sites such as Facebook every single day. 3 billion pictures. There's 3 billion people on the internet. So like literally every person on the internet takes at least one picture a day and uploads it to the site. You can play this with other industries as well. Let's take Uber. Uber digitized the rider information. Where's the car and where's the person who wants to have the car? It was deceptive. Their first product was something— was basically a black car service, so a very luxurious product. And a lot of people saw it as a luxury good. It became disruptive when Uber copied the business model from Lyft, which is UberX. So they allowed everyone to become a driver. They dematerialized car ownership. If you live in San Francisco, there's a really good chance that you don't buy a car anymore because Uber is actually cheaper and better. You demonetize it. Uber is significantly cheaper than a cab. And the ultimate goal is to put a rider into every car in every seat. Basically give you full access to transportation.
