Photonic computing? Who’da thunk it?
If the operators of Jiuzhang are telling the truth, they managed to get the photonic computer to perform an operation in 200 seconds that would take an “ordinary” supercomputer half a billion years to complete. That’s both good and bad news.
The Bad News First: I’ve long claimed that AI will never match the human brain. Now, I’m not so sure. I think I can mentally add and subtract relatively fast, but I have to use chunking to get a quotient in long division, and though it is a time saver, it does take time. So, Jiuzhang is definitely faster at arithmetic than I am. But just working out problems isn’t what matching a human brain with a computer is all about. The power of photonic computing is scary. HAL scary? (Sorry, had to get in the pun)
Even HAL had weaknesses. Otherwise, it couldn’t have been reduced to brain-stem-like thinking by the desperate astronaut Dr. Dave Bowman. HAL’s limitations lay in its immobility and in its easy-to-get-to wiring. If HAL had been mobile…well, that would have been a story Arthur C. Clarke didn’t write. So, even with fictional HAL and its 2020 AI incarnations, I haven’t considered—regardless of the late Stephen Hawking’s warning—that computers might surpass people and pose a threat to humanity. I reasoned that even with less mathematical computing ability, humans still had advantages over the HALs. People can run up steps, not swallow their chewing gum, mutter something, and catch a ball at the same time. Humans can walk over uneven terrain and talk, or take pictures, or think about investments, at the same time. No, we can’t really multitask in the office. I know multitasking is just a term for rapid consecutive actions, but we can do many tasks in near simultaneity. This Jiuzhang thing is scary because even with multiple tasks, it works so fast that its simultaneity is more simultaneous than human multitasking. Lots of sci-fi examples, of course. I’m thinking of the clever Grant Sputore and Michael Lloyd Green film I Am Mother as an example of an AI in a robot’s body. Great film in my opinion, but ultimately, a bit scary for our species.
Worse News: Am I going to be replaced by a blog-writing machine? I know, probably. OpenAI has already released a version of an artificial intelligence author that The Guardian tested. ** In an essay-writing test, GPT-3, which is OpenAI’s language generator, wrote that robots would probably harm humans. Shades of Stephen Hawking’s prediction! Shades of Sputore’s and Green’s fictional world. And, of course, shades of HAL.
Should I worry? Well, what could go wrong? OpenAI has assured me of their meticulous care in their development of AI. On their own FAQ sheet, the company writes, “We also believe that safely deploying powerful AI systems in the world will be hard to get right. In releasing the API , we are working closely with our partners to see what challenges arise when AI systems are used in the real world. This will help guide our efforts to understand how deploying future AI systems will go, and what we need to do to make sure they are safe and beneficial for everyone.” *** Yeah, and we never heard that one before. What could go wrong with humans in control? What could go wrong with big companies or big governments in control?
Scary News: And when Big Government or Big Tech gets its hands on a photonic computer and puts it into a mobile shell, where will my supposed “human superiority” go? I can’t do the calculations in 200 seconds that a supercomputer would take half a billion years to complete. I think walking up steps, not swallowing gum, muttering something, and catching a ball simultaneously would elicit a derisive essay from the AI blog-writer. Or, worse, elicit pity followed by “Don’t worry, human, I will take care of you.” Sure, just as HAL took care of the astronauts aboard The United States Spacecraft Discovery One in 2001: A Space Odyssey. If you’ve read the book or seen the Kubrick movie, you know such “care” was lethal.
So, now I’m concerned that a photonic AI will, indeed, pose a threat to humanity. At the very least it will eliminate the need for mathematicians. Who wants to wait around till they solve those so-called unsolvable problems when a computer can compute in 200 seconds a problem that would occupy other machines for the equivalent of Phanerozoic Time? I probably can’t imagine the tech that a photonic computer could generate any more than Newton could have imagined a LIGO, the device that detects gravitational waves from colliding black holes and neutron stars.
And here I am, marveling at the machine in front of me, a MAC with 2TB on a 5G WiFi. Here I am searching for just the right words to write for you, all those words somewhere swimming around in my 86 billion neurons—at least, that’s the new estimate by Dr. Suzana Herculano-Houzel. **** And my neurons swim in constantly turbulent waters: Paying bills, accounting for safety, fixing broken things, planning events and getaways, trying to help others, checking on friends and family, thinking about where I put my COVID masks, vitamin D and zinc supplements, getting supplies without catching a lethal or debilitating disease, and, as Biden says when he can't think of what he wanted to say, "You know the thing." *****
Remember that old hypothesis about a room full of monkeys all sitting in front of typewriters? (Those were mechanical devices that were at one time used to pound messages onto paper in an advanced form of putting runes on clay tablets). So, the hypothesis goes that given an indefinite time frame, the monkeys could randomly type all the great books. As Bob Newhart joked years ago, pretending to be the lab guy overseeing the monkeys, “Here’s something. This one wrote, ‘To be or not to be? That is the gzomnplat.’” Well, Jiuzhang isn’t a monkey. And if it took about 500 million years for humans to evolve a Shakespeare, it might take that photonic computer only 200 seconds to write the next Hamlet. So, if my writings pale by comparison to Shakespeare’s, how would they pale in comparison to Jiuzhang’s?
Two hundred seconds to perform a half billion years’ worth of calculations! Things are about to change, but not necessarily for the betterment of humans. Oh! Sure, at first we’ll marvel at how photonic computers make daily life easier. But then, if the sci-fi writers are prophetic, the various forms of artificial intelligence will assume we are an unnecessary blight on the world. Pessimistic? A little. Realistic? Hey, we already have an AI writing an essay that proclaims that humans will fall to its will.
Is there good news? Give me a minute… Still thinking… Oh! Self-driving cars won’t have accidents unless the photonic computers decide to run over some humans. No doubt you can think of something that is not simultaneously positive and scary. Sure, young people won’t have to go to war. Machines will do the fighting. Problem is that humans will probably not survive such a fight, becoming either indirect or direct casualties on a scale too terrible to contemplate. When photonic computers run the planet sans humans, farmlands will return to natural states, rivers won’t receive and carry anthropogenic pollutants, and many other natural processes now interrupted by human activities will go on as they have for most of the last 4.5 billion years, that is, without interference from humans.
Of course, in all of this photonic future, human consciousness won’t be around to appreciate the fruits of its inventions. In fact, we now seem to be on the cusp of inventing ourselves out of existence. But then, aren't we always on the cusp of change?
*https://www.sciencenews.org/article/new-light-based-quantum-computer-jiuzhang-supremacy
**McBreen, Kelen. AI Robot Programmed To Write An Article: The Results Are Terrifying. Newswars. 5 Dec 2020. Online at https://www.newswars.com/ai-robot-programmed-to-write-an-article-the-results-are-terrifying/
***https://openai.com/blog/openai-api/ Accessed December 6, 2020.
****Cherry, Kendra. How Many Neurons Are in the Brain? Verywellmind. 10 Apr. 2020. Online at https://www.verywellmind.com/how-many-neurons-are-in-the-brain-2794889 , Accessed December 6, 2020. The old estimate was 100 billion neurons, but Herculano-Houzel made a “brain soup,” counted, and then extrapolated, finding the old estimate to be 14 billion neurons too large a number. Still, we generally do all right with just 86 billion neurons.
*****Sorry, just have to comment. What's worse for us, a brain that thinks faster than we think or a brain that has trouble thinking?
If the operators of Jiuzhang are telling the truth, they managed to get the photonic computer to perform an operation in 200 seconds that would take an “ordinary” supercomputer half a billion years to complete. That’s both good and bad news.
The Bad News First: I’ve long claimed that AI will never match the human brain. Now, I’m not so sure. I think I can mentally add and subtract relatively fast, but I have to use chunking to get a quotient in long division, and though it is a time saver, it does take time. So, Jiuzhang is definitely faster at arithmetic than I am. But just working out problems isn’t what matching a human brain with a computer is all about. The power of photonic computing is scary. HAL scary? (Sorry, had to get in the pun)
Even HAL had weaknesses. Otherwise, it couldn’t have been reduced to brain-stem-like thinking by the desperate astronaut Dr. Dave Bowman. HAL’s limitations lay in its immobility and in its easy-to-get-to wiring. If HAL had been mobile…well, that would have been a story Arthur C. Clarke didn’t write. So, even with fictional HAL and its 2020 AI incarnations, I haven’t considered—regardless of the late Stephen Hawking’s warning—that computers might surpass people and pose a threat to humanity. I reasoned that even with less mathematical computing ability, humans still had advantages over the HALs. People can run up steps, not swallow their chewing gum, mutter something, and catch a ball at the same time. Humans can walk over uneven terrain and talk, or take pictures, or think about investments, at the same time. No, we can’t really multitask in the office. I know multitasking is just a term for rapid consecutive actions, but we can do many tasks in near simultaneity. This Jiuzhang thing is scary because even with multiple tasks, it works so fast that its simultaneity is more simultaneous than human multitasking. Lots of sci-fi examples, of course. I’m thinking of the clever Grant Sputore and Michael Lloyd Green film I Am Mother as an example of an AI in a robot’s body. Great film in my opinion, but ultimately, a bit scary for our species.
Worse News: Am I going to be replaced by a blog-writing machine? I know, probably. OpenAI has already released a version of an artificial intelligence author that The Guardian tested. ** In an essay-writing test, GPT-3, which is OpenAI’s language generator, wrote that robots would probably harm humans. Shades of Stephen Hawking’s prediction! Shades of Sputore’s and Green’s fictional world. And, of course, shades of HAL.
Should I worry? Well, what could go wrong? OpenAI has assured me of their meticulous care in their development of AI. On their own FAQ sheet, the company writes, “We also believe that safely deploying powerful AI systems in the world will be hard to get right. In releasing the API , we are working closely with our partners to see what challenges arise when AI systems are used in the real world. This will help guide our efforts to understand how deploying future AI systems will go, and what we need to do to make sure they are safe and beneficial for everyone.” *** Yeah, and we never heard that one before. What could go wrong with humans in control? What could go wrong with big companies or big governments in control?
Scary News: And when Big Government or Big Tech gets its hands on a photonic computer and puts it into a mobile shell, where will my supposed “human superiority” go? I can’t do the calculations in 200 seconds that a supercomputer would take half a billion years to complete. I think walking up steps, not swallowing gum, muttering something, and catching a ball simultaneously would elicit a derisive essay from the AI blog-writer. Or, worse, elicit pity followed by “Don’t worry, human, I will take care of you.” Sure, just as HAL took care of the astronauts aboard The United States Spacecraft Discovery One in 2001: A Space Odyssey. If you’ve read the book or seen the Kubrick movie, you know such “care” was lethal.
So, now I’m concerned that a photonic AI will, indeed, pose a threat to humanity. At the very least it will eliminate the need for mathematicians. Who wants to wait around till they solve those so-called unsolvable problems when a computer can compute in 200 seconds a problem that would occupy other machines for the equivalent of Phanerozoic Time? I probably can’t imagine the tech that a photonic computer could generate any more than Newton could have imagined a LIGO, the device that detects gravitational waves from colliding black holes and neutron stars.
And here I am, marveling at the machine in front of me, a MAC with 2TB on a 5G WiFi. Here I am searching for just the right words to write for you, all those words somewhere swimming around in my 86 billion neurons—at least, that’s the new estimate by Dr. Suzana Herculano-Houzel. **** And my neurons swim in constantly turbulent waters: Paying bills, accounting for safety, fixing broken things, planning events and getaways, trying to help others, checking on friends and family, thinking about where I put my COVID masks, vitamin D and zinc supplements, getting supplies without catching a lethal or debilitating disease, and, as Biden says when he can't think of what he wanted to say, "You know the thing." *****
Remember that old hypothesis about a room full of monkeys all sitting in front of typewriters? (Those were mechanical devices that were at one time used to pound messages onto paper in an advanced form of putting runes on clay tablets). So, the hypothesis goes that given an indefinite time frame, the monkeys could randomly type all the great books. As Bob Newhart joked years ago, pretending to be the lab guy overseeing the monkeys, “Here’s something. This one wrote, ‘To be or not to be? That is the gzomnplat.’” Well, Jiuzhang isn’t a monkey. And if it took about 500 million years for humans to evolve a Shakespeare, it might take that photonic computer only 200 seconds to write the next Hamlet. So, if my writings pale by comparison to Shakespeare’s, how would they pale in comparison to Jiuzhang’s?
Two hundred seconds to perform a half billion years’ worth of calculations! Things are about to change, but not necessarily for the betterment of humans. Oh! Sure, at first we’ll marvel at how photonic computers make daily life easier. But then, if the sci-fi writers are prophetic, the various forms of artificial intelligence will assume we are an unnecessary blight on the world. Pessimistic? A little. Realistic? Hey, we already have an AI writing an essay that proclaims that humans will fall to its will.
Is there good news? Give me a minute… Still thinking… Oh! Self-driving cars won’t have accidents unless the photonic computers decide to run over some humans. No doubt you can think of something that is not simultaneously positive and scary. Sure, young people won’t have to go to war. Machines will do the fighting. Problem is that humans will probably not survive such a fight, becoming either indirect or direct casualties on a scale too terrible to contemplate. When photonic computers run the planet sans humans, farmlands will return to natural states, rivers won’t receive and carry anthropogenic pollutants, and many other natural processes now interrupted by human activities will go on as they have for most of the last 4.5 billion years, that is, without interference from humans.
Of course, in all of this photonic future, human consciousness won’t be around to appreciate the fruits of its inventions. In fact, we now seem to be on the cusp of inventing ourselves out of existence. But then, aren't we always on the cusp of change?
*https://www.sciencenews.org/article/new-light-based-quantum-computer-jiuzhang-supremacy
**McBreen, Kelen. AI Robot Programmed To Write An Article: The Results Are Terrifying. Newswars. 5 Dec 2020. Online at https://www.newswars.com/ai-robot-programmed-to-write-an-article-the-results-are-terrifying/
***https://openai.com/blog/openai-api/ Accessed December 6, 2020.
****Cherry, Kendra. How Many Neurons Are in the Brain? Verywellmind. 10 Apr. 2020. Online at https://www.verywellmind.com/how-many-neurons-are-in-the-brain-2794889 , Accessed December 6, 2020. The old estimate was 100 billion neurons, but Herculano-Houzel made a “brain soup,” counted, and then extrapolated, finding the old estimate to be 14 billion neurons too large a number. Still, we generally do all right with just 86 billion neurons.
*****Sorry, just have to comment. What's worse for us, a brain that thinks faster than we think or a brain that has trouble thinking?
RSS Feed