Hi guys Philip English this from philipenglish.com. Welcome to the Robot Optimized Podcast where we talk about everything robotics related. For our next episode, we have Charles Simon who will talk about his book "Will Computers Revolt".
Philip English (00:14):
Hi guys. Philip English here, also known as Robo Phil, robot and enthusiasts report on the latest business and application of robotics. And my main mission is to get you guys’ robot optimized, support industry infrastructure, and innovation for the next era. I’m excited today because you’ve got Charles Simon and who’s going there and tell us a little bit about his book and we’re going to do a bit of an interview with Charles about the AI, side of technology. So, welcome Charles is very, I really appreciate your time today. It’s a perfect, so could you give us, a quick overview, I suppose, like a little bit about yourself, like a little bit about your history, if that’s okay, Charles?
Charles Simon (01:00):
Sure. I’m a long time. Silicon valley, serial entrepreneur. And, I started three of my own companies and worked at two of other startups. And I spent a couple of years working at Microsoft and doing all kinds of different things. My very first company was about computerated design of printed circuit boards. And one of the things we observed is that the way computers designed the printed circuit boards at that, in that era was seriously different from the way people did it. And people did a better job. And that intrigued me into the idea of what makes people intelligence different from artificial intelligence. And I followed through on that, a little background about myself. I’ve got a degree in electrical engineering and a master’s in computer science. And so I’ve got a little bit of academic background in the area, but the area we’re talking about is the future of computers and artificial intelligence.
Charles Simon (01:57):
And that’s so cutting edge that nobody would say I’ve got 20 years experience in that. Along the way, also I did a stint as a developer of a lot of neurodiagnostic software. And so if you get a brain injury, you might be hooked up to my software or you get carpal tunnel syndrome and all of these other things that you test testing for neural pulses. So I bring to the table a whole lot of interesting and interest in how the human brain works and how neurology works and try and map that onto the artificial intelligence world too.
Philip English (02:37):
Right. I see. So you see, you’ve got a wealth of experience there from obviously from like an academic point of view and from a business side. So you sort of like merge the two together. Like we could probably actually jump into your brain, like sits at simulator software straight away. So explain sort of like what, the, and that the brain sits in later solves.
Charles Simon (03:02):
Well, back at the entire world of artificial intelligence. Back in the 1960s, there was a divergence of artificial intelligence where there are the neural network guys and the symbolic AI guys, and they kind of went their separate ways. And since then, they’ve kind of gone back and forth and sometimes one group got a whole bunch of money and the other group faded, and now they’re back together again right now that the neural network guys now call it deep learning or deep neural networks, and they’re more or less in charge. And they have a very interesting set of solutions, but they are not related to the way your brain works. And so the idea in 1970s or early eighties was we got this great new neural network algorithm with backpropagation. If we could just put it on a big enough computer, it would be as smart as a person.
Charles Simon (04:06):
Well, in the intervening 50 years, that has proven not to be the case. And so we have to look to some different algorithms. And so I wrote that the brain simulator looking at it from the other point of view, let’s start with how neurons work and see what we can build with that. And so my electrical engineering background says, oh, well, let’s build a simulator. And, if you were building a digital simulator, you’d have basic building blocks of NAND gates. And if you were doing analog simulator, you’d have various electronic components and op amps, but in the brain simulator, the basic component is a neuron. And the way a neuron works is it accumulates ions and eventually REITs a threat reaches a threshold fires, sends that a spike down its axon to distribute more ions to other, all of the neurons it’s attached to through it.
Charles Simon (05:00):
Synapses and neurons can have lots of synapses, you know, on the order of 10,000 and your brain has got billions and billions of neurons in it. And so, but the neat thing is that neurons are so slow that a lot of the circuitry in your brain is coping with that problem and the amount of computer power that we can get simulating neurons can simulate. Now I can simulate a billion neurons on my desktop, which I couldn’t do before. So we’re getting very close to having computers that can match the power of simulated neurons. And I’ve done a lot of explorations and this is a brain simulators, a community project. So it’s all free and you can download it and you can build your own circuits. And then you will become a lot smarter about what neurons can and can’t do and see why it diverges so much from the AI backpropagation approach.
Philip English (06:04):
Wow. I want to say, let’s say so someone, like not myself, but I couldn’t really use it as a learning tool to sort of understand the subject
Charles Simon (06:13):
Like you, you, I mean, like all learning things and you can sit down in front of it and a novice can, it’s got a bunch of sample neural networks and you can say, “aha”, well, this, these are the sorts of things I can do with neurons and how you could use these to do many more advanced things as well. And so the book kind of draws the surroundings around the software to say, well, if you go down this path, it’s pretty obvious that in the next decade or so, we will have machines smarter than people. What are the implications of that? And what, what will those machines be like and how are we going to control them and what are our options? And that’s what the book is about related to the software. So they kind of work together that way.
Philip English (07:04):
Yeah, no, that makes sense. Obviously you’ve designed and built the software, so you’re the perfect expert really, to look forward and actually see that if this grows at this rate, this is what we’re going to see, like in the future. And yeah, and that I, and that leads perfectly onto the book really. So, I mean like will computers, like revolt, is the name of the book and you’ll see the siding sort of the when, why and how dangerous it is going to be. And then again, give us a brief overview of the book then. I mean, I noticed that you’ve got three main sections and then if it’s got 14 chapters and the first part seems to be sort of explaining about how it all works. And then the next section is obviously what your, what you think is going to happen like within the future?
Charles Simon (07:58):
Well, that in order to talk about making a machine that is intelligent, you need to consider the idea of what intelligence actually is. And you need to think about what it is that makes people intelligent. And this turns out to be not an easy task to say, this is an intelligent thing to do, because if you start making a list, you’ll say, you know, can read a newspaper. Well, blind people don’t read newspapers, and yet they seem to be intelligent and you could hear a symphony. And there are always these disabilities that work that are in concert with perfectly intelligent people. So you, can’t just itemize a list of say, if you can do this and this and this and this year intelligent, and if you can’t do this and this and this, you’re not intelligent because you always have this problem, but I can see some underlying abilities, like the ability to recognize patterns in an input stream.
Charles Simon (08:56):
Now I’ve made a huge abstraction jump there, but a year, you know, your senses are continuously pouring data in it, your brain and your brain is doing its best to make sense of them, to remember what are the things that are going on at whether things worked out when you made a choice of one action over another action, and then to repeat those things. So if you said, you know, a simple game of tic-tac-toe you say, well, if I saw this, this situation, I made this move and I won, or I made this move and I lost. And you, your brain builds up these memories of things that worked out and things that didn’t work out. And so intelligent behavior is doing things that worked out. And so all of this happens within the limits of what you know, and what you’re learning. And another real problem of your brain is it’s getting so much data that, that it can only really focus on a tiny percentage of it at any time and remember even less.
Charles Simon (10:01):
And so when you stop to think about what, you know, you know, a whole lot less than you think you do, you have this perception that you can remember what’s next to you, or what’s next door, or what your friends look like. But when you actually get down to drawing a picture, you have very sketchy remembrances, and your memories are very fate to get fuzzy. And so building a computer system that works in this way, we were starting with the definition of intelligence. So it got some kind of a basic definition. And then you can say, bill working with these facets, can you build a software system or a hardware system, the software first, because it’s easier. And you build a software system that does that. And the answer is yes, and it’s not that tough, but there are certain things that we want to talk about in terms of general intelligence.
Charles Simon (11:03):
And that is, well, people seem to be able to understand stuff. You can understand stuff. I can understand stuff. What does understanding mean? And to some extent, understanding is putting everything, you know, and everything your input is receiving in the context of everything else you already know. And so you’re able to merge all of this together in a multisensory sort of way, that is you hear words, or you read words, and these may mean the same thing, but they relate to abstract things, objects, or physical actions or something. So it’s not the words that are the meaning. It is an abstraction, that’s the meaning. Then you can paste words on top of that. And so you can build computer systems to do all of these things, and that’s pretty likely, and you will end up with a because it’s doing the right thing over and over, you end up with a goal directed system, because the idea of doing something that worked out versus didn’t is entirely arbitrary.
Charles Simon (12:17):
It’s based as a measurement against some goals that some program were put into place. And so if your goal is to comprehend the world and explain it to people, that’s entirely different from a goal being set of making a lot of money or taking over the world. And so we have a goal directed system that has these capabilities. Now in the last section of the book, it is, well, what will these machines actually be like? And what will they be like when they are equivalent to a three-year-old or equivalent to an adult, or unfortunately, only 10 years later after they are equivalent to an adult will be a thousand times faster than the equivalent of an adult. And so all of these things map out to what’s the future of intelligent machines. So now the final section, I map out a number of different scenarios, which kind of put them on the low levels of different likelihood.
Philip English (13:24):
Yeah, well, this is it. I had to look through the chapters and the connection that we have with robotics is obviously a lot. Robotics is it’s all about the physical world. It’s all about the sensors that are coming out every year. There’s better and better cameras as better and better laser scanners, better LIDAR. But the real intelligence we’re seeing in robotics is all the AI side. So it sets up taking the data from the modern cameras and actually using it in an efficient way to get a job done. And with that intersection of technology getting faster and faster, and AI getting faster and faster, we’re, we’re certainly going to have an exponential growth soon with of certain technologies.
Charles Simon (14:05):
Exactly. And one of the things that I’d like to add to that is robotics is a key to general intelligence, because if you start with the idea of things, a three-year-old knows that round things roll and square thing blocks can be stacked up and things like that. These are things that you might be able to put into words and explain to a computer or show in pictures and explain to a computer. But that is entirely different from the understanding you get from having played with these blocks and to set a robot with a manipulator, loose, to play with blocks, we’ll give it an entirely different level of understanding than anything you could train. And so robotics is where the general intelligence has to emerge, because it’s the only place that brings together all of these different senses.
Philip English (14:58):
Well, this is it. And this is when you get touch senses, smelling senses, tasting senses. And you know, when, when I, and understand that, and yeah, we’re certainly going to see, and they’re
Charles Simon (15:08):
Some of the real keys are the sense of time that some things have to happen before things other happened. You know, that you have to stack the blocks before they can fall down.
Philip English (15:21):
That’s been great. It’s been great. Yeah. This is really, really, like interesting. I think it’s, it’s a perfect sideline as well. Cause we will talk about products and stuff. And this is quite good to have this view.
Charles Simon (15:33):
But from a product perspective now I happen to have been very fortunate in my professional career. So in these books and brain simulator and stuff, I do not need to make any money, which is a good thing. Because if I went to somebody and said, I need a billion dollars and I’m going to build a machine that’s as good as a three-year-old. This is not a winner of a project because three year olds don’t do very much, but that is the approach you have to take. You’ve got to be able to understand what a three-year-old can understand before you can understand what an adult can understand.
Philip English (16:10):
Yeah, no, and that’s it. And then from there that you can grow. I mean, so what I was interested in is your four light scenarios. So obviously I saw that number one was like the ideal one and then there was a few others, but if you can take us through your thoughts about that.
Charles Simon (16:28):
Sure, one can eat the scenarios of what happens when machines are a lot smarter than us. And there’s an interim period where there’s where they’re smart enough to interact with us, but not so smart that we’re borrowing. So that’s the key of being really interesting where the ideal scenario is we have programmed computers just with goals that match what human goals are now. The good news is that our needs and the computer’s needs are divergent. We need land and clean air and clean water and clean food and mates and other this, this, and that, and computers don’t need anything that we need except energy. And so we may have a fight over energy, but mostly they’re going to be doing their own thing. And the real true AGI don’t need spaceships or submarines to do exploration, or, and they don’t need air conditioning to live in the desert because they can become spaceships and they can become submarines.
Charles Simon (17:39):
And so they have a different set of standards and they can go off and do their own thing and learn a bunch of stuff about the universe and hopefully share with us. Now, the scary parts are more like in the early stages, suppose a nefarious, human is running these AGI and directs them to do things that benefit this, that person or group at the expense of mankind. And that is the only scenario that has any relationship with terminators and all of science fiction, where they build machines for the purpose of taking over the world or the purpose of making themselves rich. I don’t see that as a very likely scenario because it happens in a very small window of opportunity where machines are smart enough to be useful, but not smart enough to refuse to do the work, because it doesn’t take a genius to say that setting off a nuclear war is bad for everybody.
Charles Simon (18:46):
So a computer could easily say, no, I’m not going to participate in that project. And that will be a very interesting scenario when computers start refusing to do the things we asked them to do, but that’s a separate issue. So machine going mad on its own is extremely unlikely because in order to do that, you have to set goals for the machine that are self-destructive to mankind as a whole. And I don’t see that as a very likely scenario. And, so we’ve got the mad machine and the mad man who does things. And then there is the mad what I call the mad mankind scenario. Let us imagine that humans continue to overpopulate the world at a great rate. And they do put themselves in situations where the computers can see, well, this is going to get us into trouble. We need to do something about that.
Charles Simon (19:49):
All of the things that computers might do to solve human problems are going to be things that humans are not going to like, if you know, you say they want to solve the overpopulation problem or the famine problem, you can think of lots of solutions that you’re not going to be very happy with. So the four things that you can do, there’s the pleasant scenario that the mad man scenario, the mad machine scenario, which I think is pretty unlikely and the madman kind scenario, which is a concern. And we is what really says it’s time for mankind to get its house in order, and to solve our own problems, because we won’t want machines to solve them for us.
Philip English (20:38):
That’s it. Now, perfect. No, that’s a great light synopsis of the last four. And it’s again the very interesting and there’s four different scenarios. And, I think, yeah, I mean, like, I mean, if people obviously want to get hold of the book and I know it’s on Amazon and everything is, there.
Charles Simon (20:59):
The computer bit, the name of the book is will computers revolt, and there is a website will computers, revolt.com. The name of the software is brain simulator. And there is because it’s free it’s brain sim.org.
Philip English (21:16):
Know that, that’s perfect. Thanks, Charles. And then I suppose the last question I had is that timeframe wise, obviously, like we all know about Ray Croswell and he’s 24 foot 45, like live predictions. If you, do you think it will fit along that sort of timeframe do you think it’d be longer or shorter
Charles Simon (21:34):
Shorter, but the key is that it’s not an all or nothing situation when you think of a three-year-old, it’s not obvious that that three-year-old is going to become an intelligent adult. And so everything we do, if you look at everything you don’t like about your computer systems today, it’s mostly because they don’t think they’re not very smart. And so everything we do to make our, that brings on little pieces of smartness will be so happy to get it. So the machines increasing intelligence is inevitable because all of the little components are things we want, and we’ll eventually get to machines that are smarter than us, but it will have happened so gradually that we won’t have noticed. And every step along the way, we will have enjoyed it.
Philip English (22:35):
Well, this is it. This is the benefits. I mean, I’ve recently just invested in a little light health gadget and, you know, it’s there to benefit me really, you know, and us as a species. So yeah. Hopefully if you want it there, but no, that’s great. Well, I, thanks very much for your time. The light your time, Charles, it’s very much appreciated. I mean, what I’ll do guys is I’ll send, I’ll put a link on the YouTube video, so you guys can go and get touch of Charles book. You’re gonna have a look at his brain simulator software, and then, yeah, we’ll probably, do this again, another 6, 6, 7 months time. I mean, I’m going to get a copy of the book and have a read as well. And any questions I’ll put, Charles his details so you can reach out. So thanks, Charles. Thank you very much. Fit, fit, fit, fit.
Charles Simon (23:20):
Well, thank you for the opportunity. It’s been great talking with you.
Robot Optimised Podcast #6 – Book Interview with Charles Simon
Charles Simon: https://futureai.guru/
Philip English: https://philipenglish.com/
Sponsor: Robot Center : http://www.robotcenter.co.uk