MLG 003 Inspiration

February 10th, 2017, and this is episode three. Inspiration. In this third episode, I hope to inspire you to want to be involved in the space of machine learning and artificial intelligence, or at least to keep a very close eye on it. Ai, I think, is the most important thing happening in the world right now. Period. And I hope that by the end of this podcast episode you'll see where I'm coming from. This is completely a philosophical episode. It's a lot of fun and a lot of fluff. So lest you lose credibility in me as a teacher of machine learning, just know that this is the last of its kind. We're gonna get into some gory details, of course, starting in the next episode. But this one's just for fun. And like I said, this is all philosophical. There are four sort of. Philosophical conversations that are happening around the space of AI right now, and I'm gonna talk about those four. Let's start with the most accessible, which is automation of the economy. As I pointed out in the last episode, the definition of artificial intelligence, being able to simulate any intellectual task you can imagine the types of jobs that are going to be displaced. Once AI comes in full swing, in the dawn of ai, we were, we already kind of expected maybe very menial digital data entry type jobs, tax preparation kind of jobs to go away. That's not, you know, we didn't really think of that as AI coming in and taking our jobs. We thought of that as the digitization and automation of jobs, of simple jobs. But more and more. Very high skilled intellectual jobs are being replaced as well. AI is hitting the medical industry with a vengeance. Medical diagnoses is a huge space where AI is hitting in radiology, looking at x-rays and coming up with a prognosis. While AI can do that very effectively by way of something called convolutional neural networks, surgery is now being revolutionized by ai. Very high precision surgery Robots are coming into the hospitals. And that right now they're kind of an aid for surgeons, but we don't know how long that will last before they're a replacement, at least theoretically. Of course, cars that we're seeing right now already on the streets, autonomous vehicles, Google self-driving car is in many cities already, and it's in the news, and people are discussing the implications they're discussing the economic implications, they're discussing the philosophical and safety implications. But right away you can imagine that taxi drivers. Freight truck drivers are going to be impacted by this immediately. In fact, programmers, designers and logo creators, programming web website programming, for example, there's a website called the grid.io, which is has the goal of generating a website by way of. Of artificial intelligence, if you give it some specifications, web design, Wix has a module that you can have it design your page around your content by way of artificial intelligence. And there's already a logo generator, an AI logo generator out there. I can't remember the name of it. But as you can see, even creativity is not exempt from ai. And in fact, talking about creativity, music and art is already being generated by AI and people are having trouble distinguishing between. A human generated piece of music or art and a machine learning generated piece. So there are two natural responses to this economic takeover by ai. One is fear. Are the robots gonna take my job? There's a web app on bbc.com that I'll point out in the show notes where you can look up your job and determine whether it's safe from the AI takeover. It's a lot of fun. I think, uh, nursing ranks up there is something that's safe for now, and I think it's because it involves a lot of highly unpredictable body motions where a robot. On the road, a self-driving car follows a very predictable set of rules. Well, a nurse running around and carrying clipboards and IVs and rushing to and fro and answering doctor's orders and all these things isn't something you can vary reliably predict with robotics and artificial intelligence. But one general response of the fear to the AI economic takeover is that if you look at. All the prior economic revolutions in human history, we've been okay. There was the agricultural revolution, which switched us from hunter gatherers to farmers and people adapted for the better. There was the industrial revolution that automated a lot of grunt work, physical labor, and the result of it was a stronger economy where people could focus on things that are less menial. And then there's the information age of the, you know, the invention of the internet, and everybody got on board with the internet and new jobs are created even as o others were displaced. So for example, previously you'd call a travel agent to book your flight to a destination. Well, Google Flights can get you your flight and find the best price. No problem. Those guys are out of a job. But new jobs were created in the information age, such as web developers and mobile app developers that didn't exist before that. And our economy is stronger than it was before. So every hump, every economic revolution does have a, a small impact in its time. A potentially negative con consequence on people who have the jobs that are being displaced. But as new jobs are created, people are able to adapt and our economy comes out better. On the flip side, there's an entirely different take to this economic takeover, which isn't fear, but excitement. People in this camp point to exactly what we're trying to achieve with economic automation, replacement of jobs so that we can live more comfortable lives. In the same way that the Industrial Revolution replaced very hard menial labor. While think about the more comfortable and safe lives that we're gonna be living with autonomous vehicles driving us around and delivering freight. The money saved by all this automation has to go somewhere and it's not gonna go into the pockets of these robots CEOs who are conducting meetings and pointing at graphs that are going up into the sky and shaking hands with other robots and laughing and then having beers after work. No, it's gonna go to us. The robots are built to serve us. So the optimists say. It's not gonna destroy our economy. It's gonna make life a lot more livable for us, and new jobs will be created that we can participate in, but we don't know what those might be yet. If you look at the result of any prior economic revolution, we lived more like kings after each revolution. Than ever before. And that's hopefully what's exactly about to come. We're gonna live even more kingly lives than we already live. And if you think about the way we live now in the information age with our couches and our TVs and our video games, and we can all just go out and eat, I mean, a person's net worth being negative. In a bygone era meant that they lived on the streets. A person's net worth being negative today is all too common due to student loans, and that's just because we live in this age where so much economic leniency is provided to us. And one way which the optimists think that this might pan out is something that's already being currently discussed in multiple governments, such as in Scandinavia called Universal Basic Income, UBI, where the government basically just gives us. Minimum wage. So how about that? So look, look into that UBI. So there you have it. Economic automation. The world is about to change under our feet by way of artificial intelligence, and I don't think that there's anybody who disagrees with that. This next section holds maybe a little bit less consensus than economic revolution. And this next section is about the singularity, and this part is really fun. This is where I put on my tinfoil hat and we start to go down that philosophical rabbit hole into the void. So let me define the singularity. The singularity was a term coined by Verner revenge. And championed in recent times by a guy named Ray Kurzwell. You're gonna hear his name a lot, so just get familiar with that name. Ray Kurzwell. I'll come back to him in a bit. What is the singularity in mathematics? It's kind of like you call the singularity, maybe the point of no return if you've got an exponential graph coming from the left and it's going to the right and getting bigger and bigger and going up and up and up. It goes up into the sky and in two infinity, the point at which it just clearly kind of made a break for it up into the sky is called the singularity. That's an exponential function. Now, if you zoom into the graph and you try to figure out where that point is that it made a break for it, you're not gonna be able to find it because it doesn't look like any point is substantially greater than the next by comparison to the prior difference of points. But if you kind of zoom back out and you, let's say you take a toy car and you're trying to push it along the graph from left to right, there will. Come a point where kind of the front bumper of the car seems to hit a wall, and that's the point that we call the singularity. So some say that human technology follows along this type of graph. If you look at our economic revolutions from the prior section, the agricultural revolution, the industrial revolution, and the information age, each revolution was closer to the, to the previous one, then that one was to its predecessor and higher up on the graph. And right now we're coming upon what's called the intelligence explosion, where all mental work is potentially automatable. By artificial intelligence. Now, you might be thinking, well, that doesn't sound like it's going to infinity. That just sounds like a next big hump on the graph. But here's where things change. Artificial intelligence is tailored towards specific applications, medical diagnoses, autonomous vehicles and stuff like this. The notion of artificial general intelligence where you have a. Intelligent agent, which can perform all of these tasks, generally any, any mental task generally across the board that's called artificial general Intelligence or strong ai. If you give it the task of sort of automating our economy and making our lives cushy, fantastic. If you give it the task of making a version of itself that is even better at performing that job, and then that new child robot has the one goal of improving our lives. And the other goal of making another version of itself that's even better than it, you know, this is sort of evolution, robot evolution, what's called seed ai, seed artificial intelligence. And since it doesn't take 25 years for a robot to grow old and. Come to knowledge about the world and it's gonna be way faster at mental processing and it's going to inherit the knowledge of its prior generations. You're gonna get a spiral sort of out of control of technological progress, and that is a shoot up into the sky. And that my friends, is the singularity. Some think that this is the next phase of evolution. Humans effectively stopped evolution when they invented such effective medical science. Well, the exponential graph maybe doesn't start and stop with technology. What if it goes all the way back into evolution, starting at the Cambrian explosion and must necessarily continue all the way past humanity? And therefore, it is almost our inevitable duty to create artificial intelligence so that evolution can continue. This is all a lot of fun. This is craziness, right? I promise. I'm gonna talk about real technical details of machine learning in a future podcast. I just, I want to get you inspired first. There's some really, really fun philosophy happening around why you should care about artificial intelligence. Believe it or not, you can take your side after you've heard the arguments. Alright, for the next section, are you ready for this consciousness? This one is near and dear to my heart. I love philosophy and I love the philosophy of consciousness. An interesting thing is that consciousness might be the last bastion of unsolved scientific riddles. We found the alphabet of life in the genome. We explored outer space and discovered the black hole in all these. Physical and mathematical phenomena that were previously magic. Everything was magic or religious. So for example, at one point, sickness illness, a cold was what is now considered simply bacteria was an infestation of evil spirits. Back in the day. It was magic and it was religion. Well, if you think about it, consciousness is still the domain. Of magic and religion, but surely our brains are the physical substrate of the mind. And our brains follow biochemical laws, physical laws of the universe. Therefore, can't our brains be considered to be machines? And if that's the case, are our minds the result of a machine, the byproduct of the workings of a machine? That brings into the into question the notion of free will. Do we have free will or simply are we reactionary to an environment? If the human mind is the byproduct of a machine and we're creating a computer brain, which itself is a machine, will the computer brain create as a byproduct, a computer mind? This is a very deep, deep rabbit hole of debate in the space of cognitive science. Cognitive science incorporates philosophy, neuroscience. Psychology and artificial intelligence, all these people, these philosophers coming together and debating whether or not in fact a machine, an artificial intelligence can be conscious. Let's talk about the words intelligence and consciousness. When I say that a. Machine is artificially intelligent. It is intelligent. If it is effective at performing its mental task at hand. In fact, maybe a calculator is intelligent by that definition. I think a lot of people, when they say yes, well it won't have intelligence, like humans have intelligence. Then you can say, well, okay, so human intelligence is superior to machine intelligence at present, but you can compare them. They're not apples and oranges. Intelligence is intelligence. What you're thinking about when you have a debate in mind is consciousness is the computer. Experiencing the phenomena that it's simulating or that it's seeing in the world, in the case of a self-driving car, or it's computing an algorithm, is it experiencing these things? An experience is called quaia. In the world of philosophy, of consciousness, is it experiencing these things? If you're religious, then I'm sure you already have your own conclusions about consciousness, where consciousness basically equates the soul. This is called dualism. The soul is separate from the brain. The physical substrate is separate from the byproduct in a substantial way. They're in different dimensions. That's called dualism. And one of the biggest proponents of the philosophy of dualism was Renee Descartes. While we've moved past dualism, you won't really find very many philosophers and cognitive scientists who believe in dualism. Now they believe in most likely monism, which says that the brain. Is the mind in some important way. And so how does the brain create a mind? A very prominent subfield of consciousness philosophy is called functionalism that is accepted today, and I'm gonna boil it down to this, and I may get in trouble for this, but if it walks like a duck and it talks like a duck, then internally it is experienced as being a duck. So if a human expresses intelligent action that appears to stem from consciousness, and a robot expresses similar intelligent action, then. Question mark. This segues effectively right into Alan Turing's test for Consciousness, which is called the Turing Test. If you're talking to a robot and you're talking to a human, they're both behind curtains and you can't see which one's which, and you can't tell the difference, then I think we've done it. That's called the Turing test. So this is so fun, the the discussion of can robots be conscious, can artificial intelligence. Be conscious and it warrants a whole podcast series of its own. I'm going to point out some good resources at the end of this episode, and I highly recommend digging in. They're real fun. And finally the last big philosophical throw down in the space of ai evil killer robots. So the big scare will AI inevitably rise against us. One of the most prominent books in this space is called Super Intelligence by Nick Bostrom. And an example of his thoughts goes like this, it's called the paperclip maximization algorithm. If you start utilizing AI at a paperclip corporation to maximize the production of paperclips, and you don't specify not to destroy the earth in the process, what might happen is that the AI might find a way to achieve. Consuming all the resources of the planet in order to maximize paperclip production. The general idea goes that be careful what you wish for because you may not be able to specify all the constraints. Big figures like Elon Musk, Stephen Hawking, and Bill Gates got on board with this at one time and it became really big in the press. Me personally, I don't think that there's warrant for such a scare, but I'm gonna leave it to you to come to your own conclusions by diving into the resources. So here are the resources, the book, the Singularity is near. Ray Kurzweil describes from A to Z the whole process of the singularity. And this is kind of a must read for anybody in ai. It's very fun, it's very fluffy. Um, it's, you know, not rigorously scientific, but it's very, you should read it in the space of consciousness and whether a machine can become conscious. Ray Kurzweil also. Wrote a book called How to Create A Mind. That's also a very good book. My favorite resource on Consciousness is a video series by a company called The Great Courses, and the series is called Philosophy of Mind, brain, consciousness, and Thinking Machines. And this goes over. All of the philosophy of consciousness from the beginning until now, including Renee Descartes and dualism, and including artificial intelligence. It's really, really good. I strongly recommend it. And then there's another book that I'm currently reading by Daniel Dennet, who is a prominent modern philosopher in consciousness called Consciousness Explained. And when it comes to the fear stuff, there's a book of course I mentioned called Super Intelligence by Nick Bostrom, and all of these resources, I'll put on my show notes on O-C-D-E-V-E-L. OC deve.com/podcasts/machine learning with a hyphen slash three. In the next episode, we're finally gonna get into technical details. I'm going to describe machine learning from a high level overview and how it's broken down into the subfields of supervised, unsupervised, and reinforcement learning. Hope to see you then.