Luis Perez-Breva on Artificial Intelligence Myths
2017-10-15 · Guest: Luis Perez-Breva (MIT Professor) · 51:56
Misconceptions and Reality of Artificial Intelligence
Bob Zadek interviews MIT professor Luis Perez-Breva about the misconceptions surrounding artificial intelligence. They discuss how AI is a collection of specialized tools designed to assist human productivity rather than a sentient force poised to replace human reasoning or jobs.
Topics: Artificial Intelligence, Machine Learning, Innovation, Automation, Job Displacement, Technology History
Speakers: Bob Zadek, Luis Perez-Breva, Caller (Jacob)
Introduction to Artificial Intelligence [00:26]
Bob Zadek: Good morning, everyone. Welcome to the Bob Zadek Show, the only live libertarian talk radio show on the air all weekend. The show of ideas, never the show of attitude. Thanks so much for listening this Sunday morning.
Well, we pride ourselves on being the show of ideas, and this morning’s show will be discussing perhaps the mother of all ideas. The topic: artificial intelligence. There is so much misunderstanding about the subject of artificial intelligence, and where there is misunderstanding, there is fear. Fear that robots will take over the world, the Terminator will control us all. Fear that robots will take over our jobs and we’ll be relegated only to servicing robots and machines and we will have no independent life. We have always been afraid of machines usurping our function in life. Perhaps that’s a function of human insecurity. Are those fears well-founded or are they wholly irrational?
To help us understand this somewhat complex topic, but to simplify it so we all will be experts on artificial intelligence in about 55 minutes from now, I’m happy to welcome to the show this morning Professor Luis Perez-Breva. Professor Perez-Breva is a professor at the mother of all incubator universities, MIT, over in Massachusetts. Luis is a serial innovator. He will explain what serial innovator means. He is working on the forefront of the subject of artificial intelligence, and he is feeling, as you will hear, pretty confident that a robot is not likely to take over his job or indeed take over my job or your job. Luis is here this morning to dispel the myths of artificial intelligence. Luis, welcome to the show this morning.
Luis Perez-Breva: Hi, Bob. Thank you. Thank you for having me.
Defining the “Serial Innovator” [02:52]
Bob Zadek: Oh, thank you for joining us. Now, you are a serial innovator. You have worked in several successful startups, and just to establish your creds as somebody who is a super expert on this subject, tell us about your brief experience as a serial innovator. Now, of course, you teach at MIT and you encourage other entrepreneurs to start their business and you encourage them to follow in your footsteps. So, what is—if that is the right description—a serial innovator and indeed an innovator in general in your world of artificial intelligence, machine learning, and the like?
Luis Perez-Breva: That’s a great question. So, to me personally, innovators solve real-world problems. So, there’s an obstacle out there that prevents us from reaching further. Here comes an innovator, puts something together, and that problem is no longer a problem. It’s actually something we can build from to actually create new jobs, essentially reach further. You can do that inside a company, you can do it inside academia, you can do it pretty much as an in-company as well. But the key difference is that innovators solve problems.
Bob Zadek: So you can look at the world, and to you—you being you and all other wonderful innovators out there—can look at the world and treat the world as if it’s one big suggestion box. And you open the box every night and there is somebody who has noticed a problem. Now, it may not be a person; it may be an activity, it may be human activity, it may be a business. But you thrive on—your raw material is problems, and your product is solutions. Is that a fair summary?
Luis Perez-Breva: I love the analogy with the suggestion box. The world is a suggestion box. That’s awesome, actually. Yes, that’s a fair summary.
AI Myths: Movies vs. Reality [05:11]
Bob Zadek: And I just made it up on the spot. Isn’t that awesome? So now we have this—so now we are talking about artificial intelligence. There is so much fear and loathing and misunderstanding about simply a scientific or perhaps business concept. Help us get into the—start us along the path of understanding artificial intelligence as the term is used both in the somewhat layman term and in the technical sense, the world in which you live. What is artificial intelligence and, most importantly, what is it not?
Luis Perez-Breva: Greatly put. So, first of all, artificial intelligence has become an enormous field. Enormous. But in our minds, it still looks a lot like what happens in the movies, like Terminator, as you quoted it. So, on the one hand, AI is so big that we have like 50 million different tools to tackle it. We have data science, we have big data, we have machine learning, logic, computers that play games. Each of these things is a different tiny tool. Each of them took an enormous amount of work to develop. None of them work for absolutely everything; they work for very narrow things.
But then in the movies, what you have is these modern movies in particular, what you have is these robots that seem to be able to accomplish absolutely everything because somehow they can use all those tools and even more seamlessly. Well, those robots just don’t—they just don’t exist. They exist in the movies. And if we were in the ’70s, we would think of them as good, like Data in Star Trek was or like R2-D2 was in Star Wars. But nowadays we live in an era where movies seem to depict very evil robots, and so we fear them. But it has to do with the movies, not with the field. Hopefully, this gets you the difference between the both. The movies talk about something; what we’re researching is still the very beginnings of an enormous field.
AI as a Productivity Tool [07:14]
Bob Zadek: And my first exposure in motion pictures to artificial intelligence was—I may be dating myself—2001: A Space Odyssey, with the somewhat malevolent robot with artificial intelligence was HAL. He was the name Keir Dullea communicated with HAL. And little-known fun fact: I wonder how many people out there know where HAL got its name from in the movie 2001: A Space Odyssey. Do you know, Luis? Do you know why it was called HAL?
Luis Perez-Breva: I do. I do. But I don’t want to spoil your fun.
Bob Zadek: All right. It was HAL—the letters H-A-L are one letter before I-B-M, which was then the computer giant, maybe the only computer company of any—which people knew about. So they took the letters IBM, went one letter earlier, ended up with HAL. Okay, little-known fun fact.
So, artificial intelligence is—it starts by being a machine. It does some functions. It takes away—it does with its machinery what humans do with muscle and eyesight and other senses. But a machine which is governed by or run by or controlled by artificial intelligence simply does it a lot better. Is that a fair summary? However, it has limitations, and it can’t simply replace the human. It just fills in some gaps and does some of the work a human does and does it better, and the effect is the human does a better job. So it doesn’t replace the human; it makes the human more productive and therefore more valuable. Is that a fair summary?
Luis Perez-Breva: I think it’s a fair summary. But to bring this home with an example that everyone experiences today: I’m sure today you’ve searched Google 20 times, 30 times maybe. To do the exact same amount of information searching that you did today, you would have had to go to a library, find a project, scout the library for work—that work would not be as up-to-date as the internet is. So all those things are now made possible from your computer through a text box. And sometimes you are the one that’s intelligent, searching that information and making sense of it. Google has an enormous amount of tools that are part of the AI toolkit that is not AI, but that actually help you navigate all that amount of information faster. It would be not only difficult for you to do something that Google does; it would be pointless because what you’re good at is at making sense of that information. What Google is good at is at presenting it in a way that makes it faster for you to do it. So that’s the path we’re walking towards AI. It’s a conversation with machines that make things that we could do well with a lot of training but are not value-added, so that we can then actually add the value the way we humans do it best.
Historical Fear of Machines [12:19]
Bob Zadek: Now, so obviously machines have been around forever, and machines have—putting aside certain moments of hysteria in the history of the development of machinery—machinery has been welcomed. It has improved life, made people more productive, made their lives better, made commodities cheaper, more accessible to them. And putting aside moments like the Luddite movement in England when they chopped up all the sewing machines because it was taking away jobs—putting aside those moments of irrationality—machines have generally been welcomed and embraced by society.
Today, artificial intelligence is looked upon by many people, including economists, I should add, as being representing a bit of an existential threat to our life as we have lived it forever. And the risk is taking away jobs. Now, is artificial intelligence different? Is the development of AI a different kind of job or way of life threat than all of the improvements in society going back to whenever you want to start? We can start with automation and the Ford Motor Company, we can go back earlier in time, we can go back to the knitting machines in England. Is artificial intelligence simply the next step in a very logical continuum, or is it somehow different?
Luis Perez-Breva: So, I believe AI has potential to be as amazing for us as cars were and as every single one of the innovations you’ve quoted were. What I believe people are wrong when they try to look at it as different is that they are looking at both, say, cars or sewing machines, looking at different pieces of information. So today, to think about AI, you’re looking at the press and what economists say in the press. But to look at what happened for sewing machines, you’re looking at historians. Now, when historians paint the picture of what happened, they paint a much more rational picture because there were more elements, plus they have hindsight. Today, we still don’t know the future. We only have the press, and the press is trying at as best it can to convey the fears, which is what it often does in the news. And so you’re getting a different piece of information. It’s the exact same. I bet if you could go back to the sewing machine times and read the press then, you would feel that that one is also different from every previous one there’s ever been. In that way, they are actually the same. You’re just looking at different pieces of information. It’s really hard for the press to predict the future, get it right, and then bring calm to the situation when we can’t predict the future, which we can’t.
AI vs. Human Reasoning [15:29]
Bob Zadek: Now, a machine—a machine using calculations to, I’ll say, make decisions—that may be a tad imprecise, but I suppose it’s good enough for the moment. A machine using processing a lot of data to make decisions—does it function exactly as a human would function? Is how a machine reacts to information—is it different than how a human reacts to the same information? Does a machine behave like a human, only faster, with access to more information, or does it behave differently than a human would behave doing the same activity?
Luis Perez-Breva: Very differently. Very differently. And by the way, this is a credit to how much the field has actually evolved. So at first, shortly after Turing’s work, people thought that if we could just have a computer beat a human at chess, we would have something intelligent. Well, what has happened since is that we’ve discovered how much more intelligent we are and how much more path there is to walk, which is not to say that we’re not walking it. But the objective is not to have a machine that reasons like a human; we already have humans. The objective really is to have a machine that helps us do things we can’t, that helps us look at data in ways we can’t, so that we can apply our skill set.
And so, but today, machines that think do not think at all like humans. And I’ll give you an example. You need about two examples of anything to figure out more or less what that thing is. You’re an expert as a human at coming up with themes and stories around your surroundings with very few examples. The machines that are actually we are building today require an enormous amount of examples to address a very simple task. And then they are unable to explain to you back what they did. So they think very—they hardly think, right? They’re incredibly good at spotting patterns, we’re making them increasingly better at that, but we’re nowhere close to them thinking about how to emulate humans. We actually don’t even fully understand how humans do it, and that’s even beyond AI at this point; it’s more about the study of neuroscience.
Bob Zadek: So I think what you have said, which is really important, is that we don’t—the whole field of artificial intelligence doesn’t aspire to build better humans than humans are. That’s not the goal at all. The goal is never—it’s not to replace a human in full, but rather to take many of the repetitive jobs that humans perform and do those jobs for the human so that the human becomes more productive at what it’s doing. There’s no—I guess other than for the intellectual pleasure of it—there’s no attempt to create humans who don’t call in sick, right? That’s not the goal of artificial intelligence.
Luis Perez-Breva: No, no. But plus if it were the goal—and you know, I cannot speak for absolutely every AI researcher on earth, right? But as far as I can tell, none of my colleagues wants to do that. We all want better reasoning machines that help us understand human intelligence, but not to replace it. Actually, I would go even further than that. As you know, there’ve been like what people call artificial intelligence winters, which is every 10 years since the ’50s, AI has been claimed to have failed, and so people lose interest in it, and then 10 years later it comes back. And every single time it’s failed, it’s because people assumed that what we wanted to do was to replace humans. And every single time people have tried, it’s been a miserable failure. So we have like now three or four AI winters in which we proved that every time people think of AI as replacing humans, it just does not work at all. It backfires on both the researchers and the companies that try it.
The USPS Automation Failure [19:45]
Bob Zadek: What would be an example of where an attempt was made to replace a human with artificial intelligence that failed? Can you think of an activity where it was—we spoke about playing chess earlier, but machines actually have learned to play chess quite well and actually to play Jeopardy quite well. I think recently, if I remember correctly, a machine for the first time won a game of Jeopardy against a Jeopardy superstar, if I’m not mistaken. So, is there a recent example of where an attempt was made on a reasonable scale to build a replacement human that failed in a certain activity?
Luis Perez-Breva: Well, you know, let me—let me tweak the question because we are so far from actually even thinking of a replacement human that the best examples you can find are places where people thought that resource allocation technologies or similar had evolved so much that they could easily do—you know, forget about humans. And one of the most common examples you can find in the business literature is the United States Postal Service. 20 years ago, it made a decision that it needed to cut the human workforce because there was easier ways to automate routes and to make it easier so that it didn’t need so many humans because it could solve part of that problem with computers. Well, they did so, and according to most business experts, today the USPS service missed on the flourishing of the parcel delivery business. UPS and others took over from them mostly because they lost all the human expertise along the process. So it always backfires that every time people think that through automation—and it’s not even AI, it’s automation—they can fully replace humans, they just shoot their own foot. But it’s a long-term kind of foot-shooting, if you want.
Bob Zadek: Is it a question that—to take the post office as an example—that they simply did it badly, that the task could have been done but they did a bad job at it, or is it just the nature of the task itself that made it inherently impossible?
Luis Perez-Breva: Well, you know, to fully answer that question and to be fair to the USPS system, I would need to know more about exactly what they did. But my guess, based on how I know about the field, is that there is no way for anybody that thinks that they can remove humans through automation—there is no way for them to win in the long term. Just no way. Automation allows us to free those humans that used to be doing tasks that were easily automated so that you can have them do more value-adding things. If you’re replacing a human with automation, you’ve probably—it’s probably been a long time since you’ve stopped listening to what that human has to say in terms of improvement of their own processes, and so you’ve sort of kind of forgotten or stopped imagining how you could reach further. So it has really little to do with automation and the task itself; it’s every time you use a technology to replace humans as opposed to enable humans to reach further, you’re actually misusing the technology. And that’s true whether it’s AI or something else.
Bob Zadek: This is Bob Zadek. I’m speaking with Professor Luis Perez-Breva. We are discussing artificial intelligence. Is it a threat to your job, to my job, to your life, to my life? We’ll be back in 30 really short seconds to continue our conversation with Professor Luis—with Luis. Please stay tuned. When we come back, I would love to ask Luis the question: what is artificial about artificial intelligence, and what is intelligent about artificial intelligence? So, Luis, you got 30 seconds to work on the answer. We’ll be back very shortly. Please stay tuned.
[Commercial Break]
“Artificial” vs. “Intelligence” [24:45]
Bob Zadek: Welcome back to the Bob Zadek Show, the only live libertarian talk radio show on the air all weekend. The show of ideas, not attitude, always. And the longest-running libertarian talk radio show in California. Thanks so much for listening to my conversation this morning with Professor Luis Perez-Breva. We are talking about artificial intelligence. Are we doomed to be controlled by robots, by a whole lot of big mechanical Arnold Schwarzeneggers running around telling us what to do, taking our jobs, and making us inherently useless? I doubt it. Luis will help us understand that any such fears are not well-founded. Luis, welcome back to the show this morning. And phone calls are of course welcome now. Our lines are open: 424-BOB-SHOW for questions for Luis.
Now, before the break, Luis, I presented the question as to the phrase “artificial intelligence.” What is artificial about it, and what is the intelligence component? And can there be such a thing as artificial intelligence, or is that kind of almost a contradiction?
Luis Perez-Breva: Really neat question. So, what’s artificial about it is that we are trying to get computers, robots, mechanics, systems to actually do work for us in ways that are go a little bit beyond just merely programming movement. But what we have, and what’s artificial right now, is we have an enormous set of tools. And you’ve heard about all of them. You’ve actually been using all of them for the last 20 years and you did not even know about it. So there’s machine learning, which allows us to extract patterns from a large pools of data and make models of it so that we can make more informed decisions. There is big data, there is data analytics, there is an enormous set of things actually for you to play with.
Now, for the most part, these are just tools. Intelligence is something we’ve been trying to define for years, decades, and that every time we try, we discover how much more complex it is. But essentially, the way we define it these days or the most advanced thinking is that you and I can engage in this conversation without necessarily kind of going to Wikipedia for everything we say and derive a theme from that conversation. And you and I will walk away from this smarter, each of us about how to think about this topic, and so we’ll be able to accomplish more because of this conversation we had. So it has very little to do with the actual data we had coming in; it has to do with the conversation we are actually having. So the hope or the aspiration is that with artificial intelligence, if they ever come together, you, I, and everybody will have the possibility to have this companion, if you want, that on very specific topics can engage in conversations with you and make you smarter. That’s the actual hope. But the key is that we’re trying to develop computers that can help us have these conversations. And the way you’re seeing it today is that Google—you develop a conversation, but the system is hardly intelligent. Netflix—you develop a conversation by trying to choose a movie. And many, many things today you’re already using it. So it’s just going to be to you—it’s just going to look incremental, but it will change your life because of how much further you can go. We’re nowhere there yet. So back to today, we just have a bunch of tools.
The Motion Detector Analogy [28:26]
Bob Zadek: Luis, as you were speaking earlier in the show and explaining artificial intelligence, I had a thought process. I—we all know about a very commonplace device, a motion detector, which people have in their homes as part of an alarm system. A motion detector, when it detects motion, it warns you. Now, it is reacting—it has an external stimulus, it receives information—there is motion out there—it processes the information and it does something. It tells you. Is that—if that is not artificial intelligence, albeit incredibly narrow in its function—is artificial intelligence different than that, or is it just that a billion times over and reacting to a whole lot of sources of information very quickly and making a decision to not to warn you, but to give out a conclusion? What is—how does the concept, the complex concept of artificial intelligence relate to my absurdly simple example of a motion detector in the front yard of somebody’s home?
Luis Perez-Breva: It’s a great example, actually. It relates zero. It has little to do with artificial intelligence altogether. It’s a system that’s programmed to perform a single function; it’s as dumb as it comes. But it’s a perfect example because that’s the root of today’s confusion. If you think that that’s the beginnings for artificial intelligence, then of course you see it much closer than it actually is and you see it much more threatening. But that it has nothing to do with artificial intelligence; it’s a simple line of code. You have sensors before we had computers; you could implement the same thing somewhat mechanically. There is very little connection between that and artificial intelligence in the sense of the aspiration. It’s one of the possibly earliest tools we had to even start imagining artificial intelligence. And so if you have this system at home and—but you know, let me flip this on you. If you want to be evil, that’s all you need, you know? The sensor with the with motion detection and the ability to close the door—that’s all you need. Anybody could hack into it and prevent you from walking back into your home or opening the home whenever they want. So evil needs far less than AI to accomplish their goal, right? So everybody that’s actually seeing evil everywhere and sees AI as evil should know that AI is not evil, right? Evil is something else and can be implemented by many means. But that first example you said is like—it’s not even a billion times over; it’s like a completely different thing, way above what that thing can accomplish.
Bob Zadek: So, but artificial intelligence at its core, it receives external information and it is programmed to do a certain thing, to use that information in a certain way. Whether it’s to print something or say something or react—is artificial intelligence more complex than simply receiving external information in one form or another, whether it’s sound or the equivalent of sight or the printed word, which is nothing other than information in a different form, and react to it in a certain way? Where does the—where does the intelligence of it come in? No one would think of a motion detector as being intelligent. So in the phrase “artificial intelligence,” what is the intelligence component other than it being a far more complex motion detector?
Luis Perez-Breva: Your description of intelligence is exactly the description of computing, right? Every computer takes information in and is programmed to produce an outcome, right? So any computer would do that. The incremental factor in artificial intelligence is the fact that you did not program it to do just one task, but rather you programmed it to learn, which fundamentally means engage in a conversation with that—with someone. And you already know what the definition of intelligence is, right? Which is that at your core, it’s not just about the fact that you can see things and interpret their colors or hear things and interpret the sound. The true definition of intelligence is that you can actually take that in, add it to your thoughts, and produce a response that the other party could actually bring back to you. So everyone remembers the day they went into a class, one class of so many kinds, and they realized something new from what the professor was saying and they thought that that was a very clever or intelligent remark. Everyone has at least one experience like that. And so that’s a moment of actual intelligence, both because you were able to recognize it in what someone was saying, but because you were also able to kind of make that appreciation and engage in conversation afterwards. So without the conversation aspect to it, it’s just a computer.
AI and Original Ideas [34:01]
Bob Zadek: Do people—and we have a caller, we’ll take a caller in a moment, so please stay on the line, Jacob. But one question before we go to our caller or callers: will there be a time, has there been a time—and of course I have read a passage in a book or heard a presentation by a smart person and I have said to myself, just as you have recounted, I’ve said, “I never thought of that. What a brilliant idea,” and I became smarter. Now, I don’t recall ever thinking that about an output from any machine that I interacted with. It has—I’m a lawyer—it has found a case, it has done something for me, it has done a chore that I could have done much slower and not as effectively. But it never has—I’ve never experienced an output where I have said to myself or thought, “What a smart idea that machine has given me.” Has that happened yet? Will it happen? Are machines producing smart ideas, original ideas where the audience would say, as you have said, “That person—now that machine—is really smart”? Has that happened yet?
Luis Perez-Breva: No. No, it hasn’t happened. And what will really happen is that before this ever happens, you will be able to engage with a machine in a topic that you don’t know little about and through conversation the machine will help you in a very specific kind of problems become smarter yourself. But I doubt you will actually realize how smart the machine was while that’s happening. So, and that’s what’s going to happen maybe in the next 50 years. I don’t think you’ll get to a point when yet or soon, anytime soon, where you’ll think, “Oh my god, that machine was really smart,” because that’s not what we’re trying to do. We’re still trying to think about how you can converse and have narratives and build narratives together with that machine on topics, very specific narrow topics. That’s where we are. But it’s a great way of summarizing it.
Bob Zadek: So artificial intelligence is not very intelligent. It’s just very fast.
Caller Segment: Autonomy and Control [36:34]
Bob Zadek: We have a caller, Jacob. Welcome to the show this morning. What’s on your mind?
Caller (Jacob): Yes, Professor. The Marx and Lenin—their idea was that the advent of communism would free people from the lack of autonomy recently. Facebook and other industry spokespeople have said recently that the future is autonomy. But Professor, when you look at what Congress is doing—I don’t know about the rest of the world, but Congress for all of the United States, they’re removing the roadway lane marker buttons that warn drivers that they’re drifting out of their lane. And this is for paving the way for machine drivers, if you will, the robot drivers to displace human drivers. Isn’t that actually against what Mr. Zadek is saying, that we have something to fear here, that the trend is less autonomy for humans, more autonomy for machines?
Bob Zadek: Interesting question, Jacob. So are we in danger, Luis, of having our life—taking away the ability to turn from the left lane to the middle lane to the right lane? We only will go where the machine directs us. Is that the danger? And that’s more of a sociological or governmental question than a scientific question. But Luis, your thoughts.
Luis Perez-Breva: So, it’s a super interesting question. So, the issue here is an issue I see with every technology. So as you know, and we’ve discussed this earlier, one of the things I do at MIT is we think of new technologies as superpowers and try to solve one question and one question only, which is: how do we bring this to scale so people benefit from this? So that turns out to be such a tough question that everybody glosses over. So when I think of, for instance, driverless cars, whenever they happen, I would like to signal a couple of things. First, the technology to make a car—a single one car—driverless today is incredibly expensive. Second, whenever you actually bring this technology to scale, the question will be: is Uber going to own every single car on earth? So is Uber going to go out and buy one replacement car for every driver, or is it going to offer every driver to retrofit their cars? So there are two possible futures there. In one future, the evil corporation will replace all drivers and will just make money off cars. In the other future, your car will drive on its own and you will make money out of it without you in the car. So those two futures are there, and which one you choose to imagine today is pretty much your choice. I don’t think anything anybody does ever will completely replace all human drivers, or unless we come up with a completely different ways of transport, like today you have public transportation in plenty parts of the world where the trains drive on their own. So I think that instead of actually being alarmed about whether they can replace cars, I would say or I would challenge people to think: how on earth are they going to pay to have every single car retrofit to be driving on its own, and what are they going to do with all the other drivers before you revolt, effectively? So it’s a much more complex future, and taking these technologies to scale takes an enormous amount of effort that most of these comments kind of don’t realize.
Bob Zadek: So Jacob, I guess the question is: it depends upon who happens to be in power and how much society will force you to have your decision-making taken away and governed by a machine. And we all have—Jacob, it occurs when you ask your question, it was a great question—we have experienced that a lot in our day-to-day activities. You go online with an airline and you want to do something, and the machine simply doesn’t let you do it. And you want to buy a ticket to go to a certain place, and you can’t. Now, that’s because of programming in the experience in the website of the airline, but it’s not very hard to imagine that government could turn over functions to machines and you simply are unable to do anything other than the machine, in which case you’re relegated, one can imagine in this dystopian world we envision, of moving out like Ted Kaczynski and living in a cabin in Montana and being totally off the grid. So that’s a great question, kind of scary, but I dare say one would hope that the political system would never allow that to occur. Jacob, thanks again. You always have thoughtful questions, and we appreciate you being a listener.
AI as a Different Way of Thinking [41:59]
Bob Zadek: Now, Luis, in listening to you talk about artificial intelligence, it seemed to me that the word “intelligence” is almost a perversion, because you have explained to me in answer to my question that a machine will never do anything very intelligent. It will do thing very well and very efficiently and with a low error rate. But if we think of intelligence—we all know, we all have it, I guess, we can identify the smartest person we ever met. And that person—how you answered the question was there will never be a computer that will replace what that smartest person I ever met does. So the word—it’s almost as if the word “intelligence” is a gloss; it’s trying to make A into B, because the computer will never be intelligent as we use the word in applying to humans.
Luis Perez-Breva: I like that way of thinking about it. But I go back to things Turing is credited for saying, and as you know, Turing is one of the founders of the field: that the best way to think of a computer is not as something that competes with us in terms of intelligence; it’s something that has a different way of thinking. It can take enormous amounts of data at once, look at many, many dimensions at once. We are tend to be limited to three dimensions; that’s how we move. That makes us incredibly good for certain things. So the best use we can make of computers and their special form of intelligence is to make them more and more capable in that realm so we can actually be more capable in ours. To bring a computer to our domain would mean that in order for them to be intelligent the way we are, we would need to limit them to three dimensions so that they can actually think about the world in our coordinates. And if we do that, then I don’t know—nature’s pretty smart, right? We’ve been—there’s a lot of years of evolution that has got us to where we are. We’re incredibly efficient machines, and I’m not really sure that that’s actually a very good objective to have, to have computers that just do the same things we do. We are far less expensive to build than a computer, and we can be trained in myriad ways.
Bob Zadek: And we’re but we’re more sometimes more annoying than computers are. But that’s a different topic for a different show. This is Bob Zadek. I’m having a wonderful hour speaking with Professor Luis Perez-Breva. We are talking about artificial intelligence. Is it a threat to you and I, or is it a boon to you and I? We’re going to take a one-minute break. We’ll be right back, please stay tuned as we wrap up this interesting hour on artificial intelligence. We’ll be right back.
[Commercial Break]
Closing Thoughts and Future Outlook [50:11]
Bob Zadek: This is Bob Zadek. Please share your thoughts on today’s show on the Bob Zadek Facebook page, by email to bobzadek.com, or tweet to @rzadek. I’ll try to reply to all thoughtful comments. A podcast of the show is available on iTunes, Stitcher, or SoundCloud. Subscribe or stream shows on bobzadek.com. That’s Bob Zadek, Z-A-D-E-K dot com.
[Libertarian Minute omitted]
Bob Zadek: Welcome back to the Bob Zadek Show with my guest this morning, Professor Luis Perez-Breva. We are talking about artificial intelligence. Now, regretfully, we are running out of time. Luis, before we have to end the show—and I hate having to do that—has there been—have you experienced any governmental action to slow down, to interfere with the natural growth of your field of artificial intelligence? Is the government reacting in any adverse way to the fears of people who might lose their jobs as a result?
Luis Perez-Breva: Actually, that’s an excellent question. No one has actually ever asked me that question before. Zero. Not at all. Government has never interfered with my work or to the extent of my knowledge with the work of any of my peers, because there is a disconnect between the fears that have arisen today and the reality of the field. And so there is really nothing to really fear about the work we are doing, and government has been completely out of our way, in as far as, you know, maybe giving grants or not giving grants according to research programs, but that’s just standard for this field or any field of research.
Bob Zadek: So it’s not yet—the public hasn’t yet—maybe it’s simply too new—the public hasn’t yet expressed a fear to their elected officials that their job is at risk, even though it is. But of course, it’s not their job; it’s simply the work they do, because they can do it cheaper than a machine can. And when a machine can do the job cheaper, that job is erased or eliminated. Think elevator operators, think telephone operators used to plug in those wires in the little holes in the wall and things of that nature. It’s quite natural for people to lose their jobs to technology, and those people did not join the soup lines; they simply found other ways to work. So for the minute, for the minute, there has not been any government activity to stop the growth of artificial intelligence.
Now, Luis, as maybe the last question because we are running out of time: what will be—looking out five years, a realistic time frame—what is likely to be the progression of artificial intelligence? Will it simply get more efficient and more efficient? We have about a minute left. What can we expect from the future of your field in the next five years? And we only have a minute left, Luis.
Luis Perez-Breva: So, very quickly, last thought then. That’s what I’m working on right now, following both the publication of my recent book and my work on AI. I’m very seriously and with many colleagues working on: we need to do more. We need to not just research AI; we need to start figuring out new problems we can actually solve. So my hope is that we would change the conversation away from the fears about an evil species taking over the world and towards telling the world there is an enormous amount of real-world problems today we solve and we can solve with AI and start to get that. So if I’m successful, if we are successful, in five years we will be able to reason with us about problems we can solve with AI and get an education that way and create new jobs, new kinds of jobs, the same way computers did 50 years ago before.
Bob Zadek: Luis’s book is Innovating: A Doer’s Manifesto for Starting from Hunch. Please read Luis’s book, follow Luis’s writing. This is Bob Zadek saying so long for now, as the music tells us we’re signing off for now. I’ll be back next Sunday for another hour of libertarian ideas, not attitude. Thanks so much for listening. Have a nice Sunday.