Ray Kurzweil: How to Predict the Future
In 2002, I had a conversation with Marc Maurer, president of the National Federation of the Blind. I had first worked with the NFB in 1976, helping build the first print-to-speech reading machine. Over the years, the various models of that device got smaller, but it remained a scanner-based system that required blind users to bring reading material to their desks. There is a lot of reading material that you can't bring to your desk, of course, like a sign on a wall or the bank ATM display. You could bring a menu back to your desk, but you'd probably prefer to read it in the restaurant.
For years, I had been predicting that someday, blind men and women would be able to use a pocket-size reader to read anything they wanted as they went through the day, from the labels in their clothing to the baking instructions on the back of a muffin-mix box. Now Maurer wanted to know when I thought that day would come, and I predicted that the actual hardware for sufficiently powerful digital cameras and pocket computers would be ready in four years, by the second quarter of 2006. Developing the software, I added, would also take four years, so the Kurzweil Cos. and the NFB had better get started on the project right away.
Right on schedule, the digital cameras and pocket computers with the specs that we needed became available last spring. Our software development project was completed on time, and so we introduced a new, portable reading machine for the blind this past July. Today, there are on the order of a thousand blind people reading all the print they encounter as they go through the day. Other companies have taken notice and are starting to develop competing products. As a result of our technology forecasting, however, we have a nice jump on the market.
To what do I owe this exquisite sense of timing? The simple truth is that timing is key to success as an inventor, so I've spent the past 30 years studying the rate by which information technology advances. Being an engineer, I gathered data on technology trends in different fields and built mathematical models. What I discovered is that understanding the timing of technological change is not as mysterious as most people think it is. In fact, I found that the models were surprisingly predictive, and today I have a group of 10 people at the Kurzweil Cos. helping me gather data and build these models.
The common wisdom that you can't predict the future is not all wrong. We can't predict specific things, such as whether Google's (NASDAQ:GOOG) stock will be higher or lower three years from now. But within information technology there are meaningful patterns. The evolution of information technology follows such exquisitely smooth exponential trajectories, in fact, that I can say with confidence that all information technology doubles its price performance and capacity pretty much every year. If you ask me the cost of a MIPS (million instructions per second) of computing in 2010, the cost of sequencing a base pair of DNA in 2012, or the spatial resolution of brain scanning in 2014, I can give you detailed figures and they are likely to be accurate. This has proved true for computation for more than 100 years, going back to the first data processing equipment used to automate the 1890 census.
One way to think about the patterns in information technology is to look at science, where we see other examples of remarkably predictable effects resulting from the interaction of inherently unpredictable phenomena. The laws of thermodynamics provide an example. The path of each molecule in a gas is modeled as a random walk. Yet the properties of the overall gas, made up of many chaotically interacting particles, is predictable to a high degree of precision. Technology evolution is, similarly, a chaotic system with remarkably predictable properties.
There's another wrinkle to keep in mind. When I say that information technology doubles in price performance and power each year, remember that the rate itself is expanding at an accelerated rate. It took three years to double the price performance of computing equipment in 1900, two years in 1950, and we're now doubling it every year. At today's exponential rate, doubling every year means multiplying by a thousand in 10 years and a billion in 30 years. But with the rate of acceleration continuing to grow, we will actually hit the billion mark in only 25 years. Consider the pervasive influence of information technology in today's world and multiply that by a billion in a quarter century--while we shrink the size of both electronic and mechanical technology by a factor of 100,000 in the same time frame--and you'll get some idea of how revolutionary information technology will be in the future.
All sorts of industries will be affected, beyond what we think of conventionally as computing. Take energy for example. Today, it seems like an area of grave concern, with implications from global warming to pollution to geopolitical instability. The fact that demand for energy is projected to triple within 20 years heightens our worries. Based largely on the 19th-century technology of fossil fuels, energy is not what we would consider an information technology. Not yet anyway. But when we have fully programmable nanotechnology, through which we can reorganize matter and energy at the molecular level, then we will see a revolutionary transformation.
Here's what I mean: Today we produce 14 trillion (about 1013) watts of power, 78 percent of which comes from fossil fuels. We have, however, plenty of energy in our midst. About 1017 watts of sunlight fall on the earth, or roughly 10,000 times more energy than we regularly consume. Solar panels today do a poor job of capturing this energy because they are inefficient, expensive, heavy, and difficult to integrate with building materials. Today production of solar power costs on average $8 per watt, much more than other energy sources.
The economics of solar power are poised to change dramatically, however, as a new generation of solar panels made with nanomaterials comes of age. Developed by a series of venture-backed companies eagerly jockeying to disrupt that $1.9 trillion worldwide oil industry, these innovative panels are projected to drop in price within a few years. And whether or not any of the known businesses now developing them are successful, once we have full-scale molecular nanotechnology-based manufacturing, we'll be off to the races.
At this point, energy will become an information technology dominated by massively parallel, computation-controlled molecular manufacturing processes. In 20 years, I believe solar panels will be as inexpensive as a penny per square meter. We will be able to place them on buildings and vehicles, build solar energy farms, and incorporate them into clothing for powering mobile devices. Converting 0.0003 percent of all sunlight hitting the earth, which will be feasible at that time, will let us meet 100 percent of our energy needs two decades from now. In yet another welcome change, we will be able to store the energy in nanoengineered fuel cells that will be tiny and widely distributed, a great improvement over the centralized, dangerous energy storage facilities we rely on today, such as liquid natural gas tanks.
Most discussions of global warming make no mention of the ability of nanotechnology to solve this problem within 20 years. Al Gore's movie An Inconvenient Truth never mentions nanotechnology, which in my view is a rather big oversight. The inclination to project the current rate of change into the future, what I call the "intuitive linear view," is hard-wired in us. The reality is that transformative changes happen faster and faster today. The telephone took 50 years to be adopted by a quarter of the U.S. population. The cell phone did that in thirteen years. Only five years ago, most people did not use search engines. Just three years ago we did not hear the terms "blog," "podcast," or "social network." And three years ago, people thought that it was impossible for a business to make money on Internet advertising. Today, we have Google, a company with a $157 billion market cap that does just that.
The pace of change is already so fast that the world will be a very different place by the end of the three-year planning cycle of typical business projects currently under way, let alone the six- or seven-year venture capital horizon. In my own technology projects, we bake into our development and business plans projections that call for the rapid advancement of technology, on a quarter-by-quarter basis. One pleasant result of doing this is that we often find that today's difficult tradeoffs dissolve within a short period of time. With the doubling of price performance each year in every kind of information technology, you just need to wait a short while to find that you can have your cake and eat it too.
The past is an accurate guide to the future only if we take these exponential progressions into account. But relatively few people do. We see what is right in front of us and expect that pace to continue. But a studied look at history shows that progress is exponential, not linear, and the difference is profound.
Ray Kurzweil is an inventor, the co-founder of the Kurzweil Cos., and the author of five books, including The Singularity Is Near.