The Virtual Thomas Edison
March 21, 2001 by Ray Kurzweil
As machines exceed human intelligence, will they threaten humanity? How will inventors keep up? Raymond Kurzweil lays out his vision of the future for Time Magazine’s special issue on the future.
Originally published November 30, 2000 at Time.com. Published on KurzweilAI.net March 22, 2001.
Robots–even ones more intelligent than M.I.T.’s Kismet–are coming, the inevitable result of accelerating technological revolutions. The exponential growth of computing, communications, brain scanning and brain “reverse engineering,” combined with rapid miniaturization, will bring machines that can equal or exceed human intelligence within three decades.
Is this good news? Or is this a threat to humanity’s perch of evolutionary superiority? Alarm at the specter of ceding control over the creative process to machines has catapulted the debate beyond the scientific community and into the public forum. Bill Joy, co-founder of Sun Microsystems, has written about a wide range of dangers that could arise when we no longer have our metaphorical hands “on the plug.”
Before we indulge these philosophical concerns, it’s worth exploring just how intelligent and inventive machines are evolving. A powerful paradigm for emulating the creative process in a computer is to copy the ways of nature. One particularly compelling “biologically inspired” approach is actually to simulate the process of evolution inside the computer.
Brandeis University professors Jordan Pollack and Hod Lipson recently used “genetic” algorithms to design simple robots, which were then assembled by other robots. General Electric also uses genetic algorithms, in the design of jet engines, and its simulation of evolution produces designs superior to those created by unaided human designers. Microsoft has reportedly evolved some of the software it uses to balance system resources rather than have human programmers explicitly write these codes.
Another approach is to create “neural nets”–simulated versions of neurons and their massive interconnections that, while highly simplified, are able to solve real-world design problems and come up with unexpected though still appropriate solutions. These and related methods are also used in computer programs that “automatically” create art, music and poetry. The results of emulating nature in this way can be surprisingly effective, often solving difficult engineering and other design problems. However, as a human inventor who routinely uses these techniques, I can report that I continue to feel that I am still in charge of the process; they feel like just another set of yet more powerful tools.
A Time Line
So when will this feeling of apparent control change? When will we regard machine intelligence as the true originator of something creative?
In my view, the advent of the fully creative machine will not arrive overnight but will continue to evolve in stages, as machine intelligence continues its progression up the skill ladder. The first computers were designed during World War II with pen on paper and assembled by hand with screwdrivers and wiring tools. Today a computer designer sits at a graphics terminal and specifies sets of high-level design parameters. The computer performs dozens of intermediate design stages of circuit schematics, board layouts and even the chips themselves. Then other computers actually build and assemble these components into working systems. The tasks these machines perform required highly skilled engineers and technicians just a couple of decades ago. Such “computer-assisted design” software packages are now used in every engineering discipline, as well as by architects and clothes designers. They represent the latest chapter in the story of automation, which started by amplifying the power of our muscles and in recent times has been amplifying the power of our minds. Since the Industrial Revolution was born two centuries ago with automated textile machines for the English clothing industry, we have been eliminating jobs at the bottom of the skill ladder while creating new (and, incidentally, more interesting and better paying) jobs at the top of the skill ladder. This process has progressed to the point where we are harnessing machines to assist with the creative process of creating yet better machines.
By the end of this decade, it will be possible for people without technical training to use an even more sophisticated generation of design tools to create complex electronic and mechanical systems. Many products will be designed not by research-and-development departments (at least not directly) but by professionals who understand the needs of their markets, aided by increasingly intelligent Web-research tools. Even consumers will design their own products, ranging from their clothes to their homes. We will continue to regard these machines as tools, but they will emerge as remarkably powerful amplifiers of the human creative process.
By 2020, machines will emerge as true collaborators. They will have sufficient understanding of human language and culture to monitor trends on their own. And since they will have the speed (and patience) to read most of the world’s literature and websites (albeit still not with the discernment that comes with full human intelligence), they will identify market opportunities on their own and bring them to our attention, along with their own suggested designs. We will then try out their creations either in virtual-reality simulations or as actual physical products produced by rapid prototyping machines. By this time, the line between human and machine creativity will indeed begin to blur.
Within three decades, machines will be as intelligent as humans. By 2030 the available computer hardware will exceed the memory and processing capacity of the human brain by a factor of thousands. Though raw capacity alone does not automatically provide human levels of intelligence, we will have largely completed the reverse engineering of the human brain. Powerful, biologically inspired models based on the various templates of human intelligence will be capable of simulating human thought processes and will ultimately do so at far greater speeds and with far greater overall capacity than unaided human thought.
So what would a thousand simulated scientists and engineers, each with a thousand times greater memory and each thinking at speeds at least a thousand times faster than today’s human inventors, accomplish? What would they invent? Well, for one thing, they would invent technologies that would allow them to become even more intelligent (because their intelligence is no longer of fixed capacity). They would change their own thought processes to think “bigger” and more complex thoughts-and to think them faster. When and if these “inventors” evolve to be a million times more intelligent and operate a million times faster, then in today’s terms, an hour would result in a century of progress.
The Next Question
Which, of course, brings up the issue of how we mere human inventors are going to keep up. As an inventor, I have more than a passing interest in this question. My view, however, is that these developments do not represent an alien invasion of intelligent machines. They are emerging from within our human/machine civilization, and the intelligence we are creating is both derivative of and an extension to our human intelligence. We are already placing today’s generation of intelligent machines in our bodies and brains, particularly for those with disabilities (e.g., cochlear implants for the deaf) and diseases (e.g., neural implants for Parkinson’s patients). By 2030 there will be ubiquitous use of surgery-free neural implants introduced into our brains by billions of “nanobots” (i.e., microscopic yet intelligent robots) traveling through our capillaries. These noninvasive neural implants will routinely expand our mind through direct connection with nonbiological intelligence.
These prospects will bring enormous benefits, such as vastly expanded wealth, longevity and knowledge. We will have the ability to overcome most diseases, clean up the environment and alleviate illiteracy and poverty. However, deeply intertwined with these gains will be profound new risks. New concerns will include such questions as “Who is controlling the nanobots?” and “Whom are the nanobots talking to?” For example, organizations (e.g., governments, extremist groups) could distribute trillions of undetectable nanobots that could then monitor, influence or even control our thoughts and actions. Nanobot self-replication run amuck could have the potential to create a nonbiological cancer. And as for intelligent robots, how can we be sure they will remain our faithful servants, or even our friends?
Technology has always been a double-edged sword, and we don’t have to look further than today to see both profound promise and peril. It is important to understand that these developments are not emerging from a few isolated projects but are the inevitable result of many thousands of competitive efforts. We would have to repeal free enterprise and every visage of economic competition to prevent the ongoing progression of these technologies. In the end, we will have no choice but to address the threats emerging from technology through a combination of ethical standards, technological “immune systems” and law enforcement. Although I believe the hazards are real, I am optimistic that we will ameliorate these dangers while we overcome age-old problems of human distress. The merger of humanity and its technology is the inevitable next step in the evolutionary progress of intelligence on our planet.
Within three decades machines will be as intelligent as human beings .