Stan Freeman, who wrote the screenplay for THE GOD QUESTION, was a newspaper science writer for nearly 25 years. His articles appeared in more than two dozen publications. The following article was published in The Republican of Springfield,
Mass., in 2008. For more discussion of the science behind the film, see the MEDIA page.
Human brain still superior ... but for how long?
BY STAN FREEMAN
For many techies, it's the equivalent of the Second Coming.
Singularity, as this highly anticipated future moment is known, will occur when computer intelligence reaches and then surpasses human intelligence.
At that point, smarter than those who programmed it, an intelligent machine, theoretically, will be able to rewrite its own programming to become smarter still, which will allow it to improve itself further and raise its intelligence further. And so on and so on. An ever upward spiral of digital IQ.
The shared vision of singularists, as they're called, is that the rise of super-intelligent computers will create a golden age of scientific inventions and advances, not unlike the time in which a steady stream of inventions, from the phonograph to the light bulb, flowed out of Thomas Edison's Menlo Park, N.J., laboratory following the development of electricity generation.
When will singularity arrive?
It could be quite soon, judging from the speed at which computing speed is increasing.
Perhaps the premiere U.S. futurist, Ray Kurzweil, author of "The Singularity is Near: When Humans Transcend Biology," believes both the hardware and software to recreate human intelligence will be available by the end of the 2020s, and that singularity will arrive about 2045.
"Within a quarter century, nonbiological intelligence will match the range and subtlety of human intelligence. It will then soar past it," he said. "(It) will have access to its own design and will be able to improve itself in an increasingly rapid redesign cycle."
"The nonbiological intelligence created in (2045) will be one billion times more powerful than all human intelligence today. We'll get to a point where technical progress will be so fast that ... human intelligence will be unable to follow it. That will mark the singularity," Kurzweil said.
Ben Goertzel, author of "Artificial General Intelligence," and another futurist, believes singularity could arrive even sooner.
"I believe we could potentially have a human-level (artificial intelligence, or AI) within seven to 10 years if sufficient funding or attention were paid to the task," he said.
However, there is still plenty of skepticism in the scientific community that singularity is right around the corner.
Erik Learned-Miller, a computer scientist at University of Massachusetts in Amherst who specializes in machine learning, thinks the wait for it may be much longer. "I don't think you'll see it for another century."
A fast computer is not necessarily an intelligent computer, at least not in the way humans define intelligent, he said.
"There is a common notion about computers that a bigger brain will be a better brain, that we can do more with a faster brain. There is also a notion that there is no limit to intelligence and that machines can be infinitely smart," Learned-Miller said.
However, so much of a human's understanding of the world is gained through senses that computers do not possess, he said.
"The senses are a very sophisticated way of gathering information. For instance, touch helps us learn so many things about the world and it gives us our intuition about the world. Programming the laws of physics into a computer to understand gravity isn't nearly as important as feeling the weight of something in your hand or feeling how that related to how hard it is to throw it," Learned-Miller said.
Also, humans have the benefit of a lifetime's experience to make judgments.
"A faster computer without either a source of experience or a way to process the experience that it does have is useless. We see human beings who have a narrow experience who often make incorrect inferences about new situations. If you've never been on a boat, you might fall over the first time you're on one because you haven't had that experience," he said.
"So one of the big challenges is how to get experience into a computer and to process it," Learned-Miller said.
How fast will a computer have to be to match the computing power of the human brain? The brain has about 100 billion nerve cells called neurons, each of which is capable of sending or receiving thousands of signals a second to communicate with muscles, glands and other neurons.
The estimates are that a computer will have to be capable of performing somewhere between 10 quadrillion and 100 quadrillion operations per second to match the brain, or in computerspeak, 10 to 100 petaflops. (A quadrillion is a million billions, a 1 with 15 zeros to its right.)
Computing speed has roughly doubled, on average, every 18 months in the last decade. At present, the world's fastest supercomputer is the IBM Roadrunner at Los Alamos National Laboratory. The $133 million machine attained a peak performance of 1.026 petaflops in May.
If the increase in speed continues at the present rate, it's possible that a supercomputer will reach 10 petaflops in speed by 2015 and 100 petaflops by 2025, say researchers.
However, speed is not everything. The software that will harness that speed to produce clear and effective thinking is equally as important. Efforts are underway to build sophisticated devices and optimize programs that will mimic all the human senses as well as the nuances of human thought.
"In a sense, there is greater complexity in the software," than in the hardware, said Learned-Miller.
UPDATE: Computing performance has improved faster than anticipated. As of June, 2013, Tianhe-2, a supercomputer developed by China’s National University of Defense Technology was the world’s fastest computer, having reached 33.86 petaflops.
FROM WIKIPEDIA: Kurzweil predicts the singularity to occur around 2045 whereas (science fiction writer Vernor) Vinge predicts some time before 2030. At the 2012 Singularity Summit, Stuart Armstrong did a study of artificial generalized intelligence (AGI) predictions by experts and found a wide range of predicted dates, with a median value of 2040. His own prediction on reviewing the data is that there's an 80% probability that the singularity will occur between 2017 and 2112.
Mass., in 2008. For more discussion of the science behind the film, see the MEDIA page.
Human brain still superior ... but for how long?
BY STAN FREEMAN
For many techies, it's the equivalent of the Second Coming.
Singularity, as this highly anticipated future moment is known, will occur when computer intelligence reaches and then surpasses human intelligence.
At that point, smarter than those who programmed it, an intelligent machine, theoretically, will be able to rewrite its own programming to become smarter still, which will allow it to improve itself further and raise its intelligence further. And so on and so on. An ever upward spiral of digital IQ.
The shared vision of singularists, as they're called, is that the rise of super-intelligent computers will create a golden age of scientific inventions and advances, not unlike the time in which a steady stream of inventions, from the phonograph to the light bulb, flowed out of Thomas Edison's Menlo Park, N.J., laboratory following the development of electricity generation.
When will singularity arrive?
It could be quite soon, judging from the speed at which computing speed is increasing.
Perhaps the premiere U.S. futurist, Ray Kurzweil, author of "The Singularity is Near: When Humans Transcend Biology," believes both the hardware and software to recreate human intelligence will be available by the end of the 2020s, and that singularity will arrive about 2045.
"Within a quarter century, nonbiological intelligence will match the range and subtlety of human intelligence. It will then soar past it," he said. "(It) will have access to its own design and will be able to improve itself in an increasingly rapid redesign cycle."
"The nonbiological intelligence created in (2045) will be one billion times more powerful than all human intelligence today. We'll get to a point where technical progress will be so fast that ... human intelligence will be unable to follow it. That will mark the singularity," Kurzweil said.
Ben Goertzel, author of "Artificial General Intelligence," and another futurist, believes singularity could arrive even sooner.
"I believe we could potentially have a human-level (artificial intelligence, or AI) within seven to 10 years if sufficient funding or attention were paid to the task," he said.
However, there is still plenty of skepticism in the scientific community that singularity is right around the corner.
Erik Learned-Miller, a computer scientist at University of Massachusetts in Amherst who specializes in machine learning, thinks the wait for it may be much longer. "I don't think you'll see it for another century."
A fast computer is not necessarily an intelligent computer, at least not in the way humans define intelligent, he said.
"There is a common notion about computers that a bigger brain will be a better brain, that we can do more with a faster brain. There is also a notion that there is no limit to intelligence and that machines can be infinitely smart," Learned-Miller said.
However, so much of a human's understanding of the world is gained through senses that computers do not possess, he said.
"The senses are a very sophisticated way of gathering information. For instance, touch helps us learn so many things about the world and it gives us our intuition about the world. Programming the laws of physics into a computer to understand gravity isn't nearly as important as feeling the weight of something in your hand or feeling how that related to how hard it is to throw it," Learned-Miller said.
Also, humans have the benefit of a lifetime's experience to make judgments.
"A faster computer without either a source of experience or a way to process the experience that it does have is useless. We see human beings who have a narrow experience who often make incorrect inferences about new situations. If you've never been on a boat, you might fall over the first time you're on one because you haven't had that experience," he said.
"So one of the big challenges is how to get experience into a computer and to process it," Learned-Miller said.
How fast will a computer have to be to match the computing power of the human brain? The brain has about 100 billion nerve cells called neurons, each of which is capable of sending or receiving thousands of signals a second to communicate with muscles, glands and other neurons.
The estimates are that a computer will have to be capable of performing somewhere between 10 quadrillion and 100 quadrillion operations per second to match the brain, or in computerspeak, 10 to 100 petaflops. (A quadrillion is a million billions, a 1 with 15 zeros to its right.)
Computing speed has roughly doubled, on average, every 18 months in the last decade. At present, the world's fastest supercomputer is the IBM Roadrunner at Los Alamos National Laboratory. The $133 million machine attained a peak performance of 1.026 petaflops in May.
If the increase in speed continues at the present rate, it's possible that a supercomputer will reach 10 petaflops in speed by 2015 and 100 petaflops by 2025, say researchers.
However, speed is not everything. The software that will harness that speed to produce clear and effective thinking is equally as important. Efforts are underway to build sophisticated devices and optimize programs that will mimic all the human senses as well as the nuances of human thought.
"In a sense, there is greater complexity in the software," than in the hardware, said Learned-Miller.
UPDATE: Computing performance has improved faster than anticipated. As of June, 2013, Tianhe-2, a supercomputer developed by China’s National University of Defense Technology was the world’s fastest computer, having reached 33.86 petaflops.
FROM WIKIPEDIA: Kurzweil predicts the singularity to occur around 2045 whereas (science fiction writer Vernor) Vinge predicts some time before 2030. At the 2012 Singularity Summit, Stuart Armstrong did a study of artificial generalized intelligence (AGI) predictions by experts and found a wide range of predicted dates, with a median value of 2040. His own prediction on reviewing the data is that there's an 80% probability that the singularity will occur between 2017 and 2112.