and now on book tv. we want to introduce you to baylor professor robert marks, who is also the author of non computable what you do that

artificial intelligence never. before we get into the book, professor what do you teach at baylor? well, let's let's see i only teach graduate courses that the current time i teach a course on artificial i teach a course information theory and that might sound to be boring, but if you get into things like algorithmic information, it's more exciting than. any science fiction i've ever seen or read, it's it's just astonishing. it introduces to things like things that exist that you can prove, but you can also prove they're unknowable okay. it just just an astonishing from the era of computer science. so i teach a course a multi dimensional processing and and stochastic process, which is a, which is a fancy word for noise. okay how you characterize noise and how you model noise. so we could identify you as a computer scientist? i, i am more of a computer, but yes, the things i do are in the

of computer science. absolutely. well, intelligence has become an it thing for the last ten years. so do you agree that is the future? what think that it's a new technology. and i think that it is like any other new technology simply tool and we're going to have to figure out how to incorporate it into society and it's amazing but so was electricity when it came along so it was thermonuclear energy when it came along. so artificial intelligence i think needs to be kept in perspective as a tool and it's a tool like any tool that be used for either good or evil. it depends on the person behind it and yeah, it the the outcome artificial intelligence is just astonishing and i think it's going to be astonishing in the near also so see unfortunately there's a lot of hype which is associated with the artificial intelligence like in the

terminator, skynet gets consciousness and takes over the world and all of these other things and. the premise of the book is that there are aspects of view which are non computable. so those that say artificial will someday. duplass gate what the human is done or can do is that's not going to happen. that's not good. there's certain brick walls that artificial intelligence can go through and in fact non computable you you write a i will never be or half understanding that's correct not neither will it have sentience and and interestingly these are incredibly powerful old ideas the idea of understanding goes back to a philosopher named john searle. john searle, he gave a little parable about something called, the chinese room and he chose chinese. he didn't read nor write chinese.

so he said, imagine me in a room and i'm surrounded by file cabinets and through slot in the door is is slipped a little sheet of paper. and on that paper in chinese is written a question in chinese. now doesn't read or write chinese but what he does is he gathers the question he began searching through the file cabinets he finds a match to the question. then he takes out from the cabinet a card and on the bottom of their courtesy answer to the question so he jots it down. he can copy and he refills the card and then he goes over and slips the sheet of paper through the door to the outside now from the outside, it looks like whatever is happening that room understand that chinese can read chinese it understand questions it can formulate answers. but no, the premise was is that john searle didn't know chinese. he was simply following an algorithm. so any time we see artificial intelligence that looks like it's doing something underneath there, no understanding.

so that's that's a hurdle that artificial intelligence will never go through. have you today or have ai today incorporated artificial intelligence into our lives? as i personally just generally, yes. in fact, i have some grants with the department of defense. and we do things such as reconfigurable computing, reconfigurable electronics. it turns out that the demands of the use of spectrum today require that we change electronics in real time. it used to be for spectrum, we had a very dedicated use of spectrum. so the military had a certain spectrum, a c span, which since they what are they thinking of? citizens? i don't know. i got to this. it's been confused with then but citizens band had certain citizen's band had a certain certain bandwidth, etc. but unfortunately the demand of

spectrum has been so high, especially due to cell and things like that that we're using up. and it's a natural resource. and the question is how do we adapt to it? and in a group that i work with with my colleague charlie bayliss, who is more in the electronics and in the artificial intelligence, and we have applied artificial intelligence to these reconfigured rebel computers so that we can more effectively share the spectrum with different, different people and the military, especially interested in this, because there are there used to be bands of of the spectrum were totally allocated to the military for radar. but in wisdom the government has auctioned these off for billions of dollars to people. at&t, verizon and. so all of a sudden we are forced now to share bands with the civilians and the dod is interested in how to do that and how to do that.

so that's one of the applications of artificial intelligence that we have used to. we have done in my work. but today, have you used it in your everyday life on your cell phone or getting an airplane? oh, well, absolutely. think before we talk about that, we have to what artificial intelligence is. right. and in my world, if you go to conferences they will tease apart things like computational intelligence and machine intelligence and intelligence. so they all have their nuanced definition. but if you look at the media, the i say the primary primarily used definition is anything a computer does which is astonishing. you go, wow, that was really, really cool. and so if we define artificial intelligence that then yeah, you know myself i'm still astonished by some of the things my cell phone can do just unfortunately

as a ai is a stab. we're numbed by the familiarity. i remember email when it came out. i'm old enough to remember in high school, in junior high, we were taught how to address a letter, where to put the stamp, the salutations and the and all of that stuff. and i thought, wow, email is going to make my life much easier because i'm not to have to do all of this formatting and everything, and boy, was i wrong. i mean, we're just inundated with, with, with that sort of thing. so is email artificial intelligence? well, think we're numbed by familiarity but some of the some of the more recent results are well, g.p.s. apple just came out with a camera that. does this enhancing of your photographs just astonishing clarity? and they use things artificial intelligence about something called ai. i'm almost certain the technology is called deep, convolutional networks. and so, yeah it's it's here and

it surrounds us and and it's being used all the time. and, of course, we're being monitored by places like amazon.com, google in terms of natural word process singing, the fact that we are a long time, long time ago surrendered a lot of my privacy to google. but unfortunately they give such incredible service that it's worth the tradeoff. so where is that privacy kept? i mean, you've written confidential emails etc. , where does that get stored? does it get stored or does it disappear stored? it's scored by google. now is a new and i love free enterprise for this but there's a new movement now to use something called edge computing where your your is stored like directly on your cell phone so it's not stored google and you have control your over your. your privacy and whether or not

you share some of this this information with other people or not so that hopefully is in the future and we won't have we won't have to surrender our privacy as we have every computer or book or artificial intelligence book that i've interviewed an about. they all seem to go back to ada lovelace, ada lovelace and alan turing. what was ada lovelace's contribution? our new world. ada lovelace. we're talking about the 1700 1600, i think it was a 19th century. i think. but there was this mathematician named oh, boy, it just was there. and then it went to anyway, he invented a computer and he never built the computer, but he talked about it with ada lovelace and and the ada lovelace is celebrated as the first computer programmer in history. and so she holds that

distinction. and one of the things that she said, her famous quote, is that artificial intelligence and i'm sorry, computers will never do anything creative and. that's a statement i agree with. alan turing came on later and explicitly disagreed with ada lovelace and said that no it's how you define creativity. and turing defined into called the turing test and then the turing test. i would sit in one room a would sit in the other room and would communicate by telephone by texting back and forth and that computer could fool me into thinking that it was a human being. it passed the turing test. the problem with the turing test is that it has been number one. it's been deemed a lot of people write computer software to to kind of fool you into thinking that's a human. the other thing it's looking at intelligence from across the

room you can look at something from across the room and it seems really real like flowers for example. and it looks just beautiful and it looks real and magnificent. but as you get close to it and you begin to examine it, you find that, no, there's no dirt there. you know, the leaves kind of feel funny. and that's the the way it is with artificial intelligence. so the proper of creativity was given by a professor from, rensselaer polytechnic called summer bridgewater. by the way, the guy in the 19th century was named babbage, just just resurfaced the brain. babbage was the one that created the computer that was never. so. anyway, i brings walking along with something he called the lovelace test and the lovelace for creativity in the computer that a computer or artificial intelligence, if you will is going to be creative if it does

something that the programmer or somebody with similar knowledge of the programmer can explain or is beyond the intent of the programmer. and so that's the lovelace test. and to date, i would challenge anybody that said that the lovelace test has been as been conquered. so to date has been no surpassed of the lovelace test and are not creative and are never going to be creative. that is an aspect of being non computable recently a google engineer warned that google's artificial intelligence was yes. does that doesn't that dispute what you've been saying? no, absolutely not. in fact, there's so many different ways to push back that one of them we just talked about was the searle's chinese room. even if it didn't respond back, it no understanding of what its response was. but like a lot of hype that

occurs, the media, it takes a little for the person across the room, look closely and see the the the the challenges and the shortcomings and. some of the things that came out, for example, was that the people that did this artificial intelligence called lamda had trained it to do exactly what it was doing. they hired crowd workers to add to their artificial intelligence to add up and i have a quote here. it says, the crowd are explicitly informed to in a safe, sensible, specific, interesting, grounded, informative manner. they are over 140,000 back in force on and this was captured natural word processing and guess what happened to the artificial intelligence it became sensible interesting and grounded and if informative. so so therefore no the the

artificial intelligence was trained that way. it certainly does not pass. summer brings word lovelace because this was exactly what the programmer was hoping to do. another thing which came out in the washington post is that is that some of this stuff was cherry picked. this is something that happens quite a bit in artificial intelligence you generate a bunch of output and, then you cherry pick the ones that are that are really impressive. and the washington post reported that reported an that they had they had captured some of the correspondence between this google engineer and somebody else and that yeah so a lot of this had been cherry picked a lot of it had been worked. so again it's like a lot of things. the closer look at it, the more it goes away. so artificial intelligence there, it's creativity with. certainly not. it's not demonstrated by google's lambda, which is

probably the reason that google bowed out of the situation and laid the engineer. i'm not sure what his current status is, but he they they said. come on. this is embarrassing. i think is what the what the result was professor marks what's the missing element in artificial intelligence to make it sentient, to make it innovative, to the premise of my book is it will never be there. there are certain things which are non computable. this goes back to the genius alan turing in the 1930s. now turing probably best known for helping crack the enigma code that helped when world war two for the nazi codes and but he was also mathematical genius and is the father of computer science and he was able to prove back in the 1930s that there were things which were non computable. now this was not conjecture,

this was not handling things, this was not terrible, this was a mathematical proof. and so since then there have been a number of other problems people have identified that are done computable. these are problems which computers will never, ever solve. so this begs the question are there aspects of us are non computable, right. and i think the low hanging fruit there is things such as well emotions such as love, compassion, empathy and such. but the deeper ones, as you mention, are created understanding and sentience. these look to be non computable. and the thing is is that if they were non computable in thirties, they're non computable now non computable means non computable and they are also going to be non computable in future even if we get super duper computers, it's still going to it's still going to be something which is out of the reach of artificial

intelligence and computers in general. well, you make the point non computable, you that alan turing was an atheist. why is that an important point to make. well it's it's just an interesting he was an atheist and he was really taken aback by a the death of one of his school chums um markham the name that comes to mind. i'm not sure, but markham died of tuberculosis. turing for the rest of his life really didn't. god, the reason i bring this up is because truth always survives initial intentions. so alan turing's initial intention was to prove that human beings able to be simulated by, um, by computers in fact, there was the assumption, this is the assumption that people that claim such things, this is that we are neat computers. and the irony is, is turing's

work eventually with the idea of non computable and subsequent developments contrib he did to the truth that we are non computable. so even though his initial premise was to show that everybody computable and we are computers made out of meat that no guess is work. the truth has come out. no, we are not computable. there are certain things that we do the computers will never do. you use the phrase non computable elegance, non computable elegance. okay, well there's there's so much fascinating. we began this by talking about information theory and these are things which are talked about in algorithmic information theory, which is the theory of algorithms. now computer can only do algorithms step by step procedure. and if you have something which is algorithmic, it's non computable. so the beauty about algorithmic information theory is that it

leads to all sorts of strange, incredible things. one of them is something called cheating is number cheating number is a number which exist on the computer. if it were if it were, we could solve many of the open of mathematics, some of which million dollar prizes. there are open problems else ever saw one single number. now that's unbelievable but cheating slumber can do that and we can prove that cheating number exists. but we can also prove that it is unknowable. so think that that is that is just astonishing stuff that comes of algorithmic information theory, things which exist, which are unknowable. there's lots of things which are unknowable a non computable a lot of the things that are non computable are unknowable. so so that's the reason i said there's this elegance in there and there's beautiful there's a beautiful door which opens up into the philosophical, dare i say, even the theological in

terms of in terms, of computers, pardon my naivety, but i thought all mathematical problems were solvable. oh no, no, absolutely the oldest one. i've get a little bit nerdy here, but it was in a letter from a guy named goldbach to the greatest mathematician of all kinds. all time, leonard euler. euler lived in the 18th century and. and goldbach wrote euler and said, you notice, i've noticed that even numbers can be written as the sum of two primes. so here is where i get nerdy. okay. a is a number which is only divisible by itself and one so seven. it was the first. few ones are two, three, five, seven, 11, 13, etc. so you can choose any given number. and goldbach says. i've noticed that it can be written can be written as the sum of two primes. and this has been an open problem for years.

and that's been checked up into the trillions. but in order to prove it, you have to prove it for an infinite number of numbers, right. so let me give you an example. pick a even number between one and 126. 26 is equal to. you got me here. let's see, 21 +70 21 plus five now. 21 is not prime. so why did you choose that? what was the number again, the number is 11. 11 is already prime. oh, you wanted to give you the numbers showing me 26. choose another number again. okay. choose a note. you choose a number that's even between one and 126, 26 would be equal to. 19 plus seven. both 19 and seven are prime numbers. and this has been this has been

checked again up into the trillions. and there is $1,000,000 prize for anybody that can mathema actually proved the box conjecture. it is one of the oldest conjectures that is not proven and there's just a whole slew of these in the area of mathematics that that haven't been proven. so no there are lots of open problems in mathematics that don't have a solution and may not have a solution. that's the that's the most interesting. robert marks, in your book non computable, you you give praise to rumsfeld's known unknowns phrase. yes, yes is that when it comes especially to your field. well there there are the known knowns. this is the donald rumsfeld's quote. there are the known knowns. there are the known unknowns there are the unknown unknowns. and rumsfeld didn't say that, but said there there are the unknown unknowns. let's see. i think i got that right. and this applies to problems. computer science currently. and there are some certain

things which are i believe i screwed up, but right. but there are certain things that we can do, for example, in engineering design, one of the things that we have to be careful with in engineering design is engineering ethics. we want to make sure that the final design does exactly what was what it was designed to do and no more. this is the big objection, for example, to autonomous weapons, okay, that they might go rogue well, not if you play engineering ethics and you make sure that that that you make sure that the that that the air does exactly what it was trained to do. and no, you can never get it totally certain, but i i maintain we could probably apply some of the things they do in the legal profession and get to that decision beyond a reasonable doubt, which is what you have to after convict the guy a murder right. so in order to release

artificial intelligence, for example, the military, you need to make sure that it does what it's supposed to do and does no more. unfortunately you never know. you know, unintended contingencies are going to come up. the artificial intelligence. and so that's the unknown unknowns, right? you because if you didn't know these contingencies, then you could put that in your design and actually and actually prevent that from happening. one of the things that we've shown in our research is that the unknown unknowns, the unintended contingencies increase exponentially with the increase in complexity of the system. so this is one of the barriers of so-called artificial general intelligence, which is not going to achieve human capability for many reasons. but for one of the reasons is that artificial intelligence is going to be so complex that that

it's going to run up against this increase in and dangers of doing things that it wasn't designed to do. how long have you been working in this field and what are the changes that really surprise you over the last 40 years or so? oh, that's i've been in i've been doing this. i think published my first paper 30 years ago. so i've been i've been doing this for four, three decades. and there are certain things which come out which are astonishing and i don't think this will a lot to the viewers, but in the in the nineties they came out with arabic propagation which is a way to train neural networks. that was astonishingly beautiful and it's used today in deep learning more. recently, there have been things called deep neural networks and gans generalized adversarial. these are used, for example in deep fakes. so you can go to a good place to go to. this is to look at this is the

website this person does not dot com and you're put you're put up with a with a picture and then you refresh it there's another and all of these pictures are all deepfakes that person does exist and it looks so realistic that it's just astonishing so i found that to be an astonishing application of artificial artificial intelligence and use this. i mean, you can use it again as a tool. you can use it for good like. they use it in the motion picture industry in order to do things. but you can also do it for evil reasons you know, to to put somebody's head on somebody else's body to incorporate to compromise their integrity, the sort of person they are. so yeah. so those those are some the things which i've seen just in the last i don't know, ten years, the gans and the deep convolutional neural networks very, very impressive technology

professor, you close your book non computable, you talking about the non computable you as being fearfully and wonderfully made. yes well that's actually quote from one of david saul sons. 139 and indeed it does appear and there's more and growing evidence that there are aspects of humans which are not to be computable they this is a problem which goes back to descartes you said no there's more to the mind than the brain. and so the so-called mind brain problem or, mind, body problem, ask the question, is the mind more than brain? descartes believed, yes, was fundamentally we fundamentally, i don't think that's true. and science is starting to show that. can i give one example? i have a friend michael ignore who is a neurosurgeon. okay.

if the mind is associated with a brain, every brain should have its own mind, right? kind of right. one mind per brain. so he goes in and he does something called a split brain where he goes and he severs the right and left hemisphere. now what do they do that they do that with? because epileptic will have a signal on one side of the brain that communicates the other side of the brain, which causes seizures. so a surgeon will go and do this left brain hemisphere, basically your one brain. now has become two brains. and so is there a separate mind which is associated with each of these brains? and the answer is no. the person emerges. there might be some personality bumps, but he's still a single. and so it is this of thing which is which is starting to give credence to the idea that the mind is indeed greater than the brain. i think in theology and christianity for example, when we talk about things such as the

possible, the soul and the spirit and things of that sort, and clearly i think the intent is this is more than the computable brain. so there's something happening there which is which is more than that. so that's the reason david said we are fearfully, wonderfully made in a mysterious which we may never crack. robert marks is a distinguished professor of electrical and computer science at baylor university, senior fellow at the discovery and the author. this book, non computable you what you do that artificial never will and he's been our guest on book tv. hello, everybody. peter mariveles here hoping this finds you all safe and well. on behalf of city lights, bookses