00:00:04.400 Hello, I'm up to chapter 6 now, the Jump to University.
00:00:09.200 This chapter is about a theme that runs through so much of David's work,
00:00:13.840 University. I can't get a grip on just how well understood or misunderstood or perhaps just
00:00:22.400 underappreciated this particular concept of universality is. You don't hear it much. You don't
00:00:30.480 hear of it much. I guess it is underappreciated and I think this chapter really is when it's
00:00:37.840 understood well enough, provides an avenue towards the solution of some philosophical discussions.
00:00:45.360 One of those would be the foretelling of the next chapter, chapter 7, which is called Artificial
00:00:50.240 Creativity, and there the discussion is in large part about artificial general intelligence.
00:00:56.880 But for now, what we can say is that, and I'll probably say this again in the next video,
00:01:01.680 the key issue with artificial intelligence versus artificial general intelligence is that many
00:01:09.040 people who are concerned about AI, Benaro sort are concerned it will be better than us at everything
00:01:15.440 possible. The argument goes like this, presently AI will beat humans at chess now every single time
00:01:22.800 they play, and at many other games like Go or whatever else. AI, in other words, dumb computers
00:01:31.600 can beat us at mental arithmetic, at multiplying or dividing big numbers,
00:01:36.000 or these days you can even go online and they'll do calculus for you. They'll do integration
00:01:40.320 faster than you ever could, they'll find derivatives faster than you ever could. AI can beat us at
00:01:45.200 driving now. They can probably make a better cup of coffee than a barista. So there is a long,
00:01:50.400 long list of things that can be automated. In other words, things that digital computers or
00:01:55.680 narrow AI can accomplish more efficiently than we can. Now, the argument gets a little bit slippery
00:02:02.960 because what some people then assume is, if you create a future AI, which can do every
00:02:10.160 task that we can write a program for and program a single artificial intelligence, perhaps a robot,
00:02:16.960 with all of those, then we have an AI with superhuman capabilities across all known capabilities.
00:02:23.840 And then you've got an AGI that's a super intelligence, right? No, that's wrong, that's very badly
00:02:30.640 wrong. The key capacity of a human being is solving problems we do not yet have an algorithm for.
00:02:37.840 In other words, things that cannot be automated. In the list of all AI programs, there is not
00:02:44.240 a program for tasks not yet thought of that could need automation. In other words, a creativity
00:02:50.960 algorithm. What does that have to do with anything? That would amount to a jump to universality,
00:02:57.760 the topic of this chapter. A universal algorithm would not be one that was a superset of all
00:03:03.280 currently existing known tasks, but rather something much shorter, presumably, it would be an
00:03:08.400 algorithm which could simulate any other algorithm. It wouldn't be a big long list of other algorithms,
00:03:14.240 it would be able to simulate any other, including algorithms for tasks not yet thought of.
00:03:20.800 But you wouldn't be able to specify the output of such an algorithm because the algorithm
00:03:24.320 would be creative. What it would do would be unpredictable. And that's what we are. We will come to
00:03:31.040 that kind of universality towards the end of this chapter. I am getting ahead of myself. David
00:03:37.840 begins this chapter, chapter six, and different to my other videos perhaps. I want to actually
00:03:42.560 read directly from the first few pages of this chapter with the discussion about different
00:03:47.920 languages and different types of writing. Early kinds of writing used symbols that represented
00:03:53.200 whole concepts and the vocabulary was limited. Writing came after speech, of course, so people
00:03:58.640 would have had a language presumably before they were able to write it down, sometime a long
00:04:03.360 before writing was invented. In fact, that language would have been universal. What it means for a
00:04:10.160 language to be universal is that anything and everything that can possibly be thought or imagined
00:04:16.000 can be expressed in that language. So I just want to emphasize that again,
00:04:19.680 universality is about the anything and the everything. And different kinds of systems might be
00:04:26.240 able to accomplish a wide repertoire of different tasks. But until that repertoire of tasks
00:04:31.760 becomes all encompassing, infinite, everything, all possible tasks, it doesn't have universality.
00:04:40.880 So language, natural language that we use, and natural languages that probably existed well before,
00:04:46.960 English or writing systems existed, were universal in this respect. Now, some people might want
00:04:53.440 to pull the breaks here and say, well, hold on, there's a whole bunch of things that we can't
00:04:56.480 express in language. So I just want to make a quick distinction about being able to express
00:05:03.440 something in practice and being able to express something in principle. So, Qualia, the most famous
00:05:09.360 example, I suppose, Qualia are the subjective way that things seem. And if I look out my window
00:05:15.360 right now, I can see the sky looks blue, but perhaps that blue looks different to me in the way
00:05:21.440 that it looks to you. The differences in the Qualia, our subjective experience of the blue sky,
00:05:28.480 can't be put into words, but that's not a failing of the universality of language.
00:05:33.040 That's simply a failure of our imagination. We haven't yet figured out how to conjure the
00:05:38.640 language in such a way as to describe subjective states. It doesn't mean that it's impossible
00:05:43.120 using language. It simply means we just don't know how. Problems are soluble, so one day we will
00:05:48.000 know how to do this. Now, putting all that aside, early writing systems had a finite number of
00:05:55.040 symbols that were essentially in one-to-one correspondence with a bunch of objects or concepts
00:05:59.920 that people wanted to be able to write down. These are often known as pictograms or hieroglyphics.
00:06:07.360 So a picture of the sun would represent the sun, and a picture of a crown might mean the king.
00:06:12.320 But if there was no picture in that writing system for a concept or object, then there was
00:06:19.760 nothing else that could be done except invent a whole new system. In other words, the number of
00:06:25.200 symbols would proliferate. Such a system isn't universal because at any given point in time,
00:06:33.760 there would be objects or concepts outside of that system which could not be represented by
00:06:39.600 that system. Today, Mandarin speakers and others, including people who can speak Japanese
00:06:44.720 quite often in people who can speak Korean quite often, use some traditional Chinese characters.
00:06:49.680 The traditional Chinese script numbers in the thousands, as far as I'm aware. There's thousands
00:06:56.160 of different characters if you want to learn to understand traditional Chinese. But it's rather
00:07:01.760 fixed, hence the word traditional. Korean students, for example, at schools throughout primary
00:07:06.800 school in high school, learn something like 1,800 different characters in order to be able to
00:07:11.600 understand some of the language that's used in their culture. But Korean students also learn
00:07:17.520 another alphabet, and it's called hungul. And hungul being the Korean alphabet consists of 24
00:07:23.280 characters. Now, if you don't know anything about Asian languages like Chinese, Japanese, and Korean,
00:07:28.160 you might look at the symbol side by side and think that they are all pretty much the same kind
00:07:33.680 of system, but in fact, you'd be very wrong. Korean has far more in common with the English script
00:07:38.960 than it does with the traditional Chinese script. There are 24 letters in the Korean alphabet,
00:07:43.600 there are 26 letters in the English alphabet, and you rearrange the symbols to literally represent
00:07:49.280 any object or any concept that you can possibly articulate. And if you run out of words,
00:07:54.880 you simply rearrange the symbols in order to invent another word. That's universality.
00:07:59.120 So we have alphabets where no new symbols need to be invented. We just rearranged the existing
00:08:05.760 ones to express anything we like. Or these things called logograms or pictograms, where you need
00:08:11.520 a whole new picture, a whole new symbol in order to represent a concept or object. So the former
00:08:17.920 alphabets are universal, the latter pictograms are not. Now for what it's worth is an aside,
00:08:23.120 traditional Chinese can be read, while I'm not actually being spoken by many different Asian people.
00:08:29.520 Indeed, the first invasion of Korea by the Japanese happened in 1592, and the Chinese came to
00:08:35.680 the aid of the Koreans, and at one point there were a number of points, there were meetings of the
00:08:40.800 Chinese, Japanese, and Korean. Neither of them could speak the others language, but they could
00:08:45.840 understand the traditional Chinese script, and this helped the three factions communicate.
00:08:50.960 Okay, back to the beginning of the infinity. David writes about how early scribes who invented
00:08:57.520 alphabets probably never realized the great advances they made beyond hieroglyphics. This chapter
00:09:03.680 is very much a historical overview of significant jumps that have happened to universality,
00:09:08.880 in various domains. The first of these were, well in fact, the first chronologically speaking
00:09:15.680 was the jump to universality that happened evolutionarily, namely in the DNA, and then in artifacts
00:09:22.400 of humans, including symbolic ways of representing language, through to computation and finally
00:09:28.960 our quantum computation. So let me begin my reading on page 127, where he's well into describing
00:09:36.960 the importance of alphabets. So David writes here, it is sometimes suggested that scribes
00:09:44.240 deliberately limited the use of alphabets for fear that their livelihoods would be threatened by
00:09:48.320 a system that was too easy to learn, but perhaps that is forcing too modern and interpretation on them.
00:09:53.280 I suspect that neither the opportunities nor the pitfalls of universality ever occurred to anyone
00:09:57.600 until much later in history. Those ancient innovators only ever cared about the specific
00:10:02.080 problems they were confronting, to write particular words, and in order to do that one of them
00:10:07.040 invented a rule that happened to be universal. Such an attitude may seem implausibly parochial,
00:10:13.360 but things were parochial in those days. And indeed, it seems to be recurring theme in the early
00:10:19.520 history of many fields that universality, when it was achieved, was not the primary objective.
00:10:25.760 If it was an objective at all, a small change in a system to meet a parochial purpose that just
00:10:30.320 happened to make the system universal as well. This is the jump to universality.
00:10:35.760 Just as writing dates back to the dawn of civilization, so do numerals.
00:10:39.520 Mathematicians nowadays distinguish between numbers, which are abstract entities, and numerals,
00:10:46.320 which are physical symbols that represent numbers, but numerals were discovered first.
00:10:51.040 They evolved from telemarks, or tokens, or stones, which have been used since prehistoric
00:10:55.600 times to keep track of discrete entities such as animals or days. I'll just pause there.
00:11:01.920 Don't we all remember in mathematics class throughout high school learning the
00:11:05.600 telesystem? I know I do, and it still goes on to that. It's a remarkably bizarre part of the
00:11:10.160 mathematics curriculum, learning these historical systems. David speaks not only here about telemarks,
00:11:15.440 the very earliest ways of learning how to record the number of things, as well as Roman numerals,
00:11:22.960 a completely esoteric bizarre system. I suppose the only reason for learning it now is when you
00:11:29.200 see all buildings that have the age of the building recorded, or at the end of the credits,
00:11:33.680 in a television show that tells you what year the television show or the movie was made in Roman
00:11:38.480 numerals. David goes through a lengthy exposition of the telesystem, and then of the Roman
00:11:45.520 numeral system as well. You write out the Roman numeral system never achieved universality.
00:11:51.760 It might have if vertical lists of Roman numerals have been permitted, where each new tier was
00:11:57.520 something in the vertical list like an exponent. Both Archimedes and Apollonia's got close to
00:12:04.240 inventing a universal system of recording numbers, but failed. Now you should read those stories
00:12:08.560 in the chapter. I think they're really, really interesting. David writes about how Archimedes
00:12:12.080 probably did notice the universality of his system, but chose not to allow it for bizarre reasons.
00:12:19.280 He discusses some of the reasons why, and then I'll continue on page 133. So I'm page 133.
00:12:25.520 He talks about how Archimedes chose not to make his system universal for the following,
00:12:33.040 even more speculative possibility, he says. And he writes,
00:12:37.040 The largest benefits of any universality beyond whatever parochial problem it is intended to solve
00:12:43.680 come from its being universal for further innovation. And innovation is unpredictable. So
00:12:49.520 to appreciate universality at the time of its discovery, one must either value abstract knowledge
00:12:54.240 for its own sake, or expect it to yield unforeseeable benefits. In a society that rarely
00:13:00.640 experienced change, both those attitudes would be quite unnatural. But that was reversed with
00:13:07.440 the Enlightenment, whose quintessential idea is, as I have said, that progress is both desirable
00:13:13.280 and attainable. And so, therefore, is universality. But that, as it may, with the Enlightenment,
00:13:20.320 pluralism and all arbitrary exceptions and limitations began to be regarded as inherently problematic,
00:13:25.440 and not only in science, why should the law treat an aristocrat differently from a commoner,
00:13:30.880 a slave from a master, a woman from a man. Enlightenment philosophers such as Locke set out to
00:13:36.240 free political institutions from arbitrary rules and assumptions. Others tried to derive
00:13:42.160 moral maxims from universal moral expectations, rather than merely to postulate them dogmetically.
00:13:47.760 Thus, universal explanatory theories of justice, legitimacy, and morality began to take their
00:13:55.120 place alongside universal theories of matter and motion. In all cases, universality was being
00:14:00.880 sought deliberately as a desirable feature in its own right, even a necessary feature for an idea
00:14:06.480 to be true, and not just as a means of solving a parochial problem. I'll pause there. This is
00:14:12.480 remarkable. It talks about how the tradition of criticism about the Enlightenment gifted us with
00:14:19.520 must work in concert with these jumps to universality, not only in the ways in which a universal
00:14:26.880 system of recording numbers might be, but universality across all domains, including most
00:14:33.120 importantly, morality. Morality shouldn't be applied differently to one person as compared to
00:14:38.880 another, but rather everyone in the same way. That's the universality, where everyone is treated
00:14:45.200 in the same way. universality is important as a fundamental principle upon which so many of our
00:14:54.960 intellectual domains rest, and therefore the ways in which problems will be solved. A roadblock
00:15:02.640 is having rules that have arbitrary exceptions. The arbitrary exception to any of these fundamental
00:15:11.040 universal principles are the very thing that is going to stop progress, and so when they're spotted,
00:15:16.320 that's a great way to correct an error. If you can remove an arbitrary rule, such that your
00:15:21.120 system becomes universal, you're going to make progress. I'll continue. A jump to universality,
00:15:27.920 that played an important role in the early history of the Enlightenment, was the invention of movable
00:15:32.400 type printing. movable type consisted of individual pieces of metal, each embossed with one letter
00:15:37.440 of the alphabet. Earlier forms of printing had merely streamlined writing in the same way that Roman
00:15:42.400 numeral streamline tallying. Each page was engraved on a printing plate, and thus all the symbols
00:15:47.680 on it could be copied in a single action. But given a supply of movable type, with several instances
00:15:53.760 of each letter, one does not further metal work, one merely arranges the type into words and sentences.
00:15:59.280 One does not have to know, in order to manufacture type, what the documents that will it will
00:16:03.840 eventually print are going to say, it is universal. Okay, so this is just another example of
00:16:09.040 universality in technology. We have a system of being able to print books that no longer
00:16:15.040 require us to etch into metal, or to bang a piece of metal out with the entire page. Instead,
00:16:20.560 all you need is type, movable type, little bits of metal, each of which has a letter on it,
00:16:27.200 letter printed on it, and you would have many such letters I appear to be explaining the
00:16:31.520 blindingly obvious here, but it is important to understand the significance of such a incremental
00:16:38.080 change. This incremental change of going from, going from etching out the entire page on a piece of
00:16:47.520 metal to instead, you would think it is a small change, but it is a huge change, because suddenly
00:16:52.640 you get universality, you go from the entire page being etched onto a piece of metal, which then
00:16:57.200 can be printed over and over again, to instead having individual pieces of metal with letters
00:17:03.200 on them. And now you can represent literally any possible text that could ever be written,
00:17:08.400 unlike in the first system. So I am going to skip a little in these discussions,
00:17:12.880 people talk about the programmable loom as well, which David mentions as well. So I'll skip that
00:17:19.440 and move straight onto the history of the universal computer. And so I'm up to page 135 now,
00:17:27.200 and David writes, the most momentous such technology is that of computers, on which an increasing
00:17:32.560 proportion of all technology now depends, and which also has deep theoretical and philosophical
00:17:37.360 significance. The jump to computational universality should have happened in the 1820s, when the
00:17:42.720 mathematician Charles Babbage designed a device that he called the difference engine, a mechanical
00:17:47.440 calculator which represented decimal digits by cogs, each of which could click into one of ten
00:17:52.560 positions. His original purpose was parochial to automate the production of tables of mathematical
00:17:57.280 functions such as logarithms and cosines, which were heavily used in navigation and engineering.
00:18:02.320 At that time, they were compiled by armies of clerks known as computers, which is the origin of the
00:18:07.680 word, and would notoriously ever error prone. The difference engine would make fewer errors because
00:18:14.240 the rules of arithmetic would be built into its hardware to make it print out a table of a given
00:18:19.200 function. One would program it only once, with the definition of the function in terms of simple
00:18:24.000 operations. In contrast, human computers had to use or be used by both the definition and general
00:18:31.200 rules of arithmetic thousands of times per table, each time being an opportunity for human error.
00:18:38.880 Okay, and then David mentions in that part there that the difference engine was able to calculate
00:18:47.520 and compute the results of different complex functions like cosines and logarithms.
00:18:55.840 This is based on the work of somebody called Brooke Taylor, and if anyone does mathematics
00:18:59.440 or university, they have to go through Taylor series. Taylor series is a rather remarkable concept
00:19:05.520 where you can take any old function that you like. It doesn't involve complex numbers. Any old
00:19:11.200 function that you like, and you can represent it as a series of multiplications and additions.
00:19:16.480 And like I say, anyone who's done undergraduate maths has to go through many exercises and
00:19:20.160 repeatedly doing these Taylor series expansions as they're called. Okay, so I'm going to skip
00:19:26.320 a next short part and David writes about how the difference engine was improved by someone making
00:19:33.600 it programmable, Charles Babbage made it programmable. And this new machine was called the
00:19:41.120 analytical engine and David writes, Babbage and his colleague, the mathematician Ada,
00:19:48.080 Countess of Lovelace, knew that this machine, the analytical engine, would be capable of computing
00:19:53.920 anything that human computers could. And that this included more than just arithmetic,
00:19:58.400 it could do algebra, play chess, compose music, process, images, and so on. It would be what is
00:20:03.520 today called a universal classical computer. I shall explain the significance of the
00:20:08.080 provides a classical in chapter 11 when I discuss quantum computers which operate at a still
00:20:12.640 higher level of universality. Neither they nor anyone else for ever a century afterwards
00:20:18.240 imagined today's most common uses of computations such as the internet word processing database
00:20:22.960 searching in games. But another important application they did foresee was making scientific
00:20:28.000 predictions. The analytic engine would be the analytical engine would be a universal simulator
00:20:34.000 able to predict the behavior to any design accuracy of any physical object given the relevant
00:20:40.560 laws of physics. This is the universality that I mentioned in chapter 3 through which physical
00:20:45.840 objects that are unlike each other and dominated by different laws of physics such as brains and
00:20:50.080 quasars can exhibit the same mathematical relationships. I'll pause there. Here's one of the most
00:20:56.320 philosophically significant issues that David raises in both the beginning of infinity and the
00:21:03.760 fabric of reality. It is one of the most contentious philosophical claims even though it's true
00:21:08.800 that has been made over the years by many different people from babage and love lace here
00:21:14.960 through to cheering, through to David Deutsch and others. And that is this idea here
00:21:23.120 said in the sentence that I've just read that David wrote where the analytical engine would be a
00:21:29.440 universal simulator. Okay so able to predict the behavior and model simulate the behavior of
00:21:38.480 any physical system. In other words, Ben includes biological systems, biological systems are a
00:21:45.760 kind of physical system. The brain is a kind of physical system. Now unless you are willing to
00:21:53.280 believe in supernatural explanations, then this tells us that computers are able to simulate,
00:22:02.240 not even electronic computers, things like the analytical engine, can simulate the functioning
00:22:08.000 of a human brain. And so they should be able to be conscious. Whatever consciousness is,
00:22:15.680 it must be a property of brains and brains are made of atoms which are performing computations
00:22:22.640 or at least the neurons are performing the computations. I'm going to skip a little bit more
00:22:28.000 and David writes up to page 137 now. Babage and love lace also thought about one application
00:22:34.320 of universal computers that has not been achieved to this day, namely so-called artificial intelligence
00:22:40.320 AI. Since human brains are physical objects obeying the laws of physics and since the analytical
00:22:45.440 engine is a universal simulator, it could be programmed to think in every sense that humans can,
00:22:50.560 albeit very slowly and requiring an impractically vast number of punched cards. Nevertheless,
00:22:56.720 Babage and love lace denied that it could. Love lace argued that the analytical engine has no
00:23:02.720 pretensions whatever to originate anything. It can do whatever we know, how to order it to perform.
00:23:10.320 It can follow analysis, but it has no power of anticipating any analytical relations or truths.
00:23:16.720 The mathematician and computer pioneer Alan Turing later called this mistake,
00:23:20.320 Lady Love Laces objection. It was not computational universality that love lace failed to
00:23:25.360 appreciate, but the universality of the laws of physics. Science at the time had almost no
00:23:31.120 knowledge of the physics of the brain. Also, Darwin's theory of evolution had not yet been published,
00:23:36.000 and supernatural accounts of the nature of human beings were still prevalent.
00:23:39.600 Today, there is less mitigation for the minority of scientists and philosophers who still believe
00:23:44.160 that AI is unobtainable. For instance, the philosopher John Searle has placed the AI project
00:23:50.400 in the following historical perspective. For centuries, some people have tried to explain the
00:23:55.600 mind in mechanical terms, using similes and metaphors based on the most complex machines of the day.
00:24:01.440 First, the brain was supposed to be like an immensely complicated set of gears and levers.
00:24:05.760 Then it was hydraulic pipes. Then steam engines, then telephonic changes,
00:24:10.240 and now the computers are our most impressive technology. Brains are said to be computers.
00:24:15.440 But this is still no more than a metaphor say so, and there is no more reason to expect the
00:24:20.080 brain to be a computer than a steam engine. This is even though an unpopular opinion above
00:24:33.200 philosophers, scientists, academics, people engaged in the debate today,
00:24:38.240 I would still say it's the overwhelming majority of the man on the street who would say that
00:24:44.080 indeed it's impossible for a computer to think in the way that people can. And that assuming
00:24:51.920 that what a brain is doing is a kind of computation is still a metaphor, is best thought of as a
00:25:01.200 metaphor. I think there was a bit of an article that's been going around recently
00:25:06.960 to this effect. Sometimes it appears in popular science articles and that kind of thing that
00:25:10.800 the brain is doing something special that the brain can't possibly be a computer.
00:25:16.000 Okay, so let me reread that last bit and continue.
00:25:21.760 This is still no more than a metaphor says so, and there is no more reason to expect the brain
00:25:26.080 to be a computer than a steam engine. But there is. A steam engine is not a universal simulator,
00:25:32.480 but a computer is, so expecting it to be able to do whatever neurons can is not a metaphor.
00:25:38.320 It is a known and proven property of the laws of physics as best we know them.
00:25:43.440 And as it happens, hydraulic pipes could also be made into a universal classical computer,
00:25:47.120 and so could gears and levers as Babidge showed. By the way, we should just
00:25:51.440 acknowledge the fact that when David writes, it is a known and proven property of the laws of
00:25:59.520 physics as best we know them. He proved that. Okay, so that's his proof. He proved that the laws
00:26:04.720 of physics in terms of quantum, the quantum laws of physics.
00:26:22.640 Okay, we just need to acknowledge something before we do move on. I'll just reread this section
00:26:27.360 again where David writes, but hey, computer is, so expecting it to be able to do whatever neurons
00:26:33.520 can is not a metaphor. It is a known and proven property of the laws of physics
00:26:38.480 pause there. Important to realize that known and proven property of the laws of physics
00:26:44.800 was proven by David Deutsch. So that's the seminal 1985 paper where he shows how the field of
00:26:54.640 computation, what was previously the mathematics of computation, if you like, was in fact a branch
00:27:00.800 of physics, because computers are made out of atoms, atoms obey the laws of quantum theory,
00:27:06.560 so therefore there can be such a thing as a quantum theory of computation.
00:27:12.720 So let's continue. David writes, ironically, lady love laces objection has almost the same logic
00:27:19.840 as Douglas Hofsted as argument for reductionism, so you chapter five, yet Hofsted is one of
00:27:25.280 today's foremost proponents of the possibility of AI. That is because both of them share the mistake
00:27:30.560 and premise that low-level computational steps cannot possibly add up to a higher level i that
00:27:36.080 affects anything. The difference between them is that they chose opposite horns of the dilemma that
00:27:41.680 that poses. Love laces chose the false conclusion that AI is impossible,
00:27:46.960 well, Hofsted chose the false conclusion that no such i can exist. Okay, so this is,
00:27:52.800 and this is also, I'll just pause there, this is me speaking, this is also I think the mistake
00:28:00.080 of some people who are fascinated by meditation and introspection that on introspection,
00:28:09.120 they can't find an eye. I first encountered this idea with the brilliant philosopher David Hume,
00:28:16.560 and David Hume talks about that whenever he reflects on his internal state, he can find nothing
00:28:22.160 but a perception. He has no perception of the eye. He just has perceptions of everything else.
00:28:28.000 Now it is true that if you reflect on yourself, you don't find an eye, but the inability to find
00:28:35.920 an eye isn't the same as demonstrating the eye doesn't exist. It's exactly the same error as
00:28:42.320 searching for God everywhere, and assuming on the basis of searching your entire bedroom,
00:28:47.200 your entire world, the universe as you know it, and being able to find God that that is somehow
00:28:51.760 a refutation of the fact that God exists. Now there might be other reasons to reject God,
00:28:56.400 but you can't form an exhaustive search. So your introspection and being able to
00:29:02.000 unable to find the eye is not a refutation of the fact that the eye exists. There are better reasons,
00:29:08.960 a better explanation for knowing the eye exists, mainly that the human being is a universal
00:29:15.680 explainer. Now when you are introspecting and you are attempting deliberately not to explain anything,
00:29:23.040 and this is what meditators do, to calm their mind and to clear their mind of all questions,
00:29:27.600 and to not contemplate anything, and to notice thoughts merely as thoughts, and not as part of
00:29:33.120 oneself, they then conclude on that basis that because there is no sensation of the self,
00:29:38.240 that therefore there is no self. That's a mistake. There's a whole bunch of things we might not have
00:29:42.880 a sensation of, but nonetheless still exist. We exist because we are universal explainers and we can
00:29:51.840 explain anything. It's the potential of what we can do that gives us ourselves. We'll come to the
00:29:58.160 remarkable explanation of what a person is towards the end of this chapter as well. It's a theme
00:30:04.560 that runs throughout the book. We are intimately not tired with the production of knowledge.
00:30:10.000 We intimately tired with the physics of computation. Human being is the nexus of these two things
00:30:16.560 in in large part. Let me continue. David writes, because of Babi just failure,
00:30:23.040 either to build the universal computer, or to persuade others to do so, and entire century would
00:30:27.920 pass before the first one was built. During that time, what happened was more like the ancient
00:30:32.240 history of universality. Although calculating machines similar to the difference engine
00:30:36.720 were being built by others, even before Babi should give an up, the analytical engine was almost
00:30:40.720 entirely ignored, even by mathematicians. In 1936, Turing developed the definitive theory of
00:30:48.320 universal classical computers. His motivation was not to build such a computer, but only to use
00:30:53.440 the theory abstractly to study the nature of mathematical proof. And when the first universal
00:30:58.240 computers were built a few years later, it was, again, not out of any special intention to implement
00:31:03.600 universality. They were built in Britain and the United States during the Second World War for
00:31:07.600 specific wartime applications. Now, I'm going to skip a bit about the history of Iniac, the first
00:31:17.120 universal computer that was built using vacuum tubes. And David writes, after skipping a bit,
00:31:25.120 David writes, it is a remarkable fact that in that sense, that is to say ignoring issues of
00:31:30.160 speed, memory capacity, and input-output devices, the human computers of old, the steam,
00:31:36.960 powered analytical engine with all its literal bells and whistles, the room-sized vacuum computers
00:31:42.720 of the Second World War, and present-day supercomputers, and your mobile phone I'll just add,
00:31:48.400 and anything else that pretty much has a computer in it, all David writes, all have an identical
00:31:54.640 repertoire of computations. In other words, they can do precisely the same thing,
00:32:00.400 some will just do it faster, some will do it slower. David writes, another thing that they all
00:32:06.560 have in common is that they are all digital. They operate on information in the form of discrete
00:32:10.960 values of physical variables, such as electronic switches being on or off or cogs being at one
00:32:15.760 of ten positions, the alternative analog computers, such as slide rules, which represent information
00:32:22.000 as continuous physical variables, were once ubiquitous, but a hardly ever used today.
00:32:27.440 That is because a modern digital computer can be programmed, imitate any of them,
00:32:31.200 and to outperform them in almost any application. The jumped university in digital computers
00:32:35.920 has left analog computation behind. That was inevitable because there is no such thing as a universal
00:32:41.440 analog computer. I'll pause there and just remark that the next paragraph talks about error
00:32:48.480 correction. As I read through it, just keep in mind the parallel we have with computation
00:32:55.920 and with epistemology. I'd like to highlight the idea that as you listen to this, we think about
00:33:03.040 how critical rationalism, the popularian view of knowledge, is very much like a digital epistemology.
00:33:11.440 On the other hand, opponents of critical rationalism, all opponents, as far as I can tell,
00:33:18.880 reject the core idea of popularian critical rationalism, that true epistemology,
00:33:26.320 and try to substitute an analog epistemology. Let's see why. I'll just read this section from
00:33:34.640 David and he writes, having just talked about analog versus digital computers in the difference,
00:33:42.800 it's impossible to have an analog universal analog computer he says.
00:33:47.920 Because that is because of the need for error correction. During lengthy computations,
00:33:53.280 the accumulation of errors due to things like imperfectly constructed components,
00:33:57.920 thermal fluctuations, and random outside influences makes analog computers wander off the intended
00:34:04.320 computational path. This may sound like a minor or parochial consideration, but it is quite the
00:34:10.480 opposite. Without error correction, all information processing and hence all knowledge creation
00:34:16.720 is necessarily bounded. Error correction is the beginning of infinity. So pause there, that is so
00:34:25.120 very important. Popularian epistemology is digital. It's digital precisely because it gives you
00:34:31.760 our binary choice between the things we know and the things we reject and no longer know or we
00:34:38.160 know a false. Knowledge has that form. Knowledge allows for error correction. Variant, variant
00:34:47.040 epistemologies, different epistemologies, reject that core idea, the binary black and white distinction
00:34:53.840 between truth and falsity, the known and the unknown, and attempt to substitute in its place
00:34:59.440 degrees of knowability, degrees of truth, degrees of belief, degrees of justification.
00:35:06.800 This is where we get probabilistic ideas in epistemology like Bayesianism,
00:35:11.840 that you can have a certain amount of confidence, but this is false. This is an analog attempt
00:35:18.400 at constructing epistemology. All of them can be rejected because what we need is not a focus
00:35:24.160 upon how close we can get to the truth or how close we can get to anywhere, but rather a way of
00:35:30.000 correcting errors. A way of identifying what the errors are and eliminating them and keeping what
00:35:36.880 remains. That's digital. I'll keep going. Okay, so I'll keep going, but I'll just skip past a
00:35:44.720 lengthy part about exactly what the problems with analog counting would be. So I'm going to skip
00:35:50.800 past that and David goes through, imagine a goat herd who's trying to tally up the number of
00:35:56.240 goats in the herd by using an analog system versus a digital system. The errors accumulate in the
00:36:02.160 analog system, and eventually you end up with the wrong number compared to the digital system
00:36:06.000 where you can correct the error. And so we get back to computation itself, David writes,
00:36:15.680 so all universal computers are digital and all use error correction with the same basic logic
00:36:20.320 that I have described, though with many different implementations. Thus, Babidge's computers are
00:36:25.280 signed on a 10 different meanings to the whole continuum of angles at which a cogwheel might
00:36:29.760 be oriented, making the representation digital in that way allowed the cogs to carry out
00:36:34.640 error correction automatically. After each step, any slight drift in the orientation of the wheel
00:36:38.960 away from its 10 ideal positions would immediately be corrected back to the nearest one as it clicked
00:36:43.760 into place. Assigning meanings to the whole continuum of angles would nominally have allowed each
00:36:49.120 wheel to carry infinitely more information, but in reality, information that cannot be reliably
00:36:54.800 retrieved is not really being stored. I'm skipping a paragraph, and David writes,
00:37:01.760 because of the necessity for error correction, all jumps to universality occur in digital systems.
00:37:08.880 It is why spoken languages build words out of a finite set of elementary sounds, speech would
00:37:14.320 not be intelligible if it were analog, it would not be possible to repeat, nor even to remember
00:37:20.160 what anyone had said. Nor, therefore, does it matter that universal writing systems cannot
00:37:24.960 perfectly represent analog information such as tones of voice. Nothing can represent those perfectly.
00:37:30.480 For the same reason, the sounds themselves can represent only a finite number of possible meanings.
00:37:34.880 For example, humans can distinguish between only about seven different sound volumes.
00:37:40.000 This is roughly reflected in standard musical notation, which has approximately seven different
00:37:44.320 symbols for loudness, such as PMF, F, and so on. For the same reason, speakers can only intend
00:37:50.880 a finite number of possible meanings with each utterance. Another striking connection between
00:37:55.840 all those diverse jumps to universality is that they all happened on earth. In fact,
00:38:00.320 all known jumps to universality happened under the auspices of human beings, except one,
00:38:05.600 which I have not mentioned yet, and from which all the others historically emerged,
00:38:10.640 it happened during the early evolution of life. Genes, in present-day organisms,
00:38:15.760 replicate themselves by a complicated and very indirect chemical route. In most species,
00:38:21.040 they act as templates for forming stretches of a similar molecule RNA. Those then act as programs,
00:38:27.280 which direct the synthesis of the body's constituent chemicals, especially enzymes, which are
00:38:31.520 catalysts. A catalyst is a kind of constructor. It promotes a change among other chemicals while
00:38:37.440 remaining unchanged itself. Those catalysts in turn control all the chemical production and
00:38:42.720 regulatory functions of an organism, and hence define the organism itself, crucially including a
00:38:47.840 process that makes a copy of the DNA. How that intricate mechanism evolved is not essential here,
00:38:53.600 but for definiteness, let me sketch one possibility. About four billion years ago,
00:38:59.360 soon after the surface of the earth had cooled sufficiently for liquid water to condense,
00:39:03.440 the oceans were being churned by volcanoes, meteor impact storms and much stronger tides in today's
00:39:08.000 because the moon was closer. They were also highly active chemically, with many kinds of molecules
00:39:12.800 being continually formed and transformed, some spontaneously and some by catalysts. One such
00:39:17.120 catalyst happened to catalyze the formation of some of the very kinds of molecules from which it
00:39:21.440 itself was formed. That catalyst was not alive, but it was the first hint of life. It had not yet
00:39:27.600 evolved to be a world-targeted catalyst, so it accelerated the production of some other chemicals,
00:39:32.560 including variants of itself. Those that were best at promoting their own production,
00:39:36.960 and inhibiting their own destruction, relative to other variants, became more numerous.
00:39:41.120 They too promoted the construction of variants of themselves and so evolution continued.
00:39:45.360 Gradually, the ability of these catalysts to promote their own production became robust and
00:39:49.600 specific enough for it to be worth calling them replicators. Evolution produced replicators
00:39:54.080 that caused themselves to be replicated ever faster and more reliably. Different replicators began
00:39:58.800 to join forces and groups, each of whose members specialized in causing one part of a complex
00:40:03.200 web of chemical reactions, whose net effect was to construct more copies of the entire group.
00:40:07.200 Such a group was a rudimentary organism. At that point, life was at a stage roughly analogous
00:40:13.680 to that of non-universal printing, or Roman numerals. It was no longer a case of each replicator
00:40:19.040 for itself, but there was still no universal system being customized or programmed to produce
00:40:24.720 specific substances. OK, pause there, so David's getting to how DNA itself evolved.
00:40:31.120 This is a momentous event in the history of the universe. Once we have replicating, once we have
00:40:38.160 DNA that becomes universal, we are able to have evolution of all the forms of life that we now have
00:40:44.640 on earth. I'll continue reading David Wright's genes are replicators that can be interpreted
00:40:50.000 as instructions in a genetic code. Genomes are groups of genes that are dependent upon each other
00:40:54.560 for replication. The process of copying a genome is called a living organism. Thus, the
00:40:59.600 genetic code is also a language for specifying organisms. At some point, the system switched to
00:41:04.880 replicators made of DNA, which is more stable than RNA, and therefore more suitable for storing
00:41:10.080 large amounts of information. The familiarity of what happened next can obscure how remarkable
00:41:16.800 and mysterious it was. Initially, the genetic code and the mechanism that interpreted were both
00:41:22.160 evolving along with everything else in the organisms. But there came a moment when the code stopped
00:41:27.680 evolving. Yet the organisms continued to do so. At that moment, the system was coding for nothing
00:41:33.600 more complex than primitive single-celled creatures, yet virtually all subsequent organisms on earth
00:41:38.080 to this day have not only been based on DNA replicators, but have used exactly the same alphabet
00:41:44.080 of bases grouped into three base words, with only a small variation in the meaning of those words.
00:41:50.480 That means that, considered as a language for specifying organisms, the genetic code has
00:41:55.200 displayed phenomenal reach. It evolved only to specify organisms with no nervous systems,
00:42:00.880 no ability to move or exert forces, no internal organs, and no sent organs, whose lifestyle
00:42:06.240 consisted of little more than synthesizing their own structural constituents and then dividing
00:42:10.000 into two. And yet the same language today specifies the hardware and soft work for countless
00:42:15.120 multicellular behaviors that had no close analog in those organisms, such as running and flying
00:42:20.240 and breathing and mating and recognizing predators and prey. It also specifies engineering
00:42:24.000 structures such as wings and teeth and nanotechnology such as immune systems and even a brain that
00:42:27.680 is capable of explaining quasars, designing other organisms from scratch and wondering why it
00:42:32.080 exists. I'll pause there. Now, I don't know if this is a well understood area of biology. I don't
00:42:38.560 think that it is exactly how the DNA molecule is able to contain information sufficient to describe,
00:42:50.560 not only bacteria, but everything through little human being and a dinosaur and flying things
00:42:57.600 and swimming things. It's rather remarkable. It has this kind of phenomenal reach, as David said,
00:43:03.600 in order to, in the space of all possible organisms that exist, it seems to have
00:43:11.520 universality. All organisms that can be built out of the organic material that you and I are
00:43:17.520 built out of. Okay, but then David talks about how early on on Planet Earth, the evolution of DNA
00:43:27.040 led to the evolution of simple bacteria, archaea, these things that lived in warm pools on the
00:43:34.800 surface of the Earth most likely. But then for something like a billion years, a billion years,
00:43:40.800 nothing happened. Literally just about nothing happened. The surface of the planet was covered
00:43:47.040 in bacteria, but no multicellular organisms evolved. No plants, no fish, nothing. Why? Why was
00:43:56.960 there no selection pressure? Why was this universality that was within the code, the DNA code? What
00:44:03.840 was it not exploited? David writes on this problem, reach always has an explanation, but this time
00:44:13.600 to the best of my knowledge, the explanation is not yet known. If the reason for the jump in reach
00:44:18.560 was that it was a jump to universality, what was the universality? The genetic code is presumably
00:44:24.560 not universal for specifying forms of life, since it relies on specific types of chemical
00:44:28.800 such as proteins. Of course, it's not universal for science fiction writers like to write about
00:44:37.120 the possibility of having organisms made of silicon. Well, DNA contains no silicon,
00:44:41.840 so copy universal for all forms of life, because presumably you'll be able to make, or perhaps,
00:44:47.360 perhaps you can make a life form out of silicon. But in particular, as he says there,
00:44:51.920 our forms of life are made of proteins, so the DNA codes for different kinds of proteins,
00:44:58.640 or the genes code for a specific protein, and so unless the life form is made out of proteins,
00:45:05.440 then the DNA isn't universal for all forms of life, but maybe it's universal for all forms
00:45:10.080 of life made out of proteins. David writes, could it be a universal constructor, perhaps?
00:45:17.520 It does manage to build with inorganic material sometimes, such as the calcium phosphate in bones,
00:45:24.000 or the magnetite in the navigation system inside a pigeon's brain. Biotechnologists are already
00:45:28.960 using it to manufacture hydrogen and to extract uranium from seawater. It can also program
00:45:33.920 organisms to build constructors outside their bodies, birds build nests, beavers build dams.
00:45:40.160 Perhaps it would be possible to specify in the genetic code an organism whose life cycle includes
00:45:44.240 building a nuclear-powered spaceship, or perhaps not. I guess it has some lesser and yet
00:45:50.800 some lesser and not yet understood universality. In 1994, the computer scientists and molecular
00:45:58.560 biologists Leonard Alderman designed and built a computer composed of DNA together with some simple
00:46:03.280 enzymes and demonstrated it was capable of performing some sophisticated computations. At the time,
00:46:08.480 Alderman's DNA computer was arguably the fastest computer in the world. Further, it was clear that
00:46:12.480 a universal classical computer could be made in a similar way. Hence, we know that whatever
00:46:17.200 that other universality of DNA system was, the universality of computation had also been
00:46:23.120 inherited in it for a billion years without ever being used until Alderman used it.
00:46:30.880 The mysterious universality of DNA, as a constructor, may have been the first universality to
00:46:36.080 exist. But of all the different forms of universality, the most significant physically is the
00:46:41.120 characteristic universality of people, namely, that they are universal explainers which makes them
00:46:48.720 universal constructors as well. Well, we have to pause there, don't we? This is amazing and this is
00:46:57.840 I think one of the least understood, but most significant parts of the entire book. It is a
00:47:06.640 discovery of David Deutsch, a philosophical and scientific discovery, a tomorrow discovery as well.
00:47:12.800 I mean, it stretches across all domains. This explanation of what a human is, what a person is,
00:47:20.400 I should say, what a person that's called humans of people, but all people share this,
00:47:24.400 the universality of people. And there might be other things as well beyond this,
00:47:29.040 but this is absolutely crucial to appreciate that any alien intelligence that we find,
00:47:34.240 any so-called supernatural being, any artificial general intelligence that we find,
00:47:42.320 it must have this capacity, this capacity to be a universal explainer,
00:47:49.680 which means that anything that can be explained, can be explained by us.
00:47:54.640 And given David's discovery of how computation is a part of physics, and you can have a physical
00:48:07.920 system, namely a quantum theater, that can simulate the behavior of any other system,
00:48:14.880 and we, as universal explainers, can perform, we, we, we can do exactly what a universal computer can
00:48:23.280 do. We can act as a Turing machine, and a Turing machine can simulate any physical,
00:48:29.760 so the universal Turing machine can simulate any physical system. So this is a poorly understood
00:48:36.240 area of physics, philosophy, morality, computation, this idea of what people are,
00:48:46.880 people are universal explainers. A universal explainer is another level of abstraction above
00:48:52.960 the universal computer. We can do what a universal computer can do. A universal computer doesn't
00:48:58.960 really need to do much more than make marks on paper, and we can do that. We can make marks
00:49:03.760 on paper, so we can do what a universal computer can do. Beyond that, we can also explain stuff,
00:49:10.880 and an explanation is a kind of computation, but you can be a universal computer without being
00:49:17.040 a universal explainer, okay? The, the computer on which I'm recording this right now is not a
00:49:22.320 universal explainer, but it's a universal computer. Now, what's needed for that extra step,
00:49:27.840 we don't know, okay? It's something to do with creativity. It's probably something to do with free
00:49:32.720 will. These things are tied up possibly intimately together in ways that we don't fully understand
00:49:38.000 yet, but this is a massive advance in the philosophy of what's going on, that people are universal
00:49:45.040 explainers. So let me just read this last paragraph. Again, I'll read it in its entirety. The
00:49:52.960 mysterious universality of DNA as a constructor may have been the first universality to exist,
00:49:58.560 but of all the different forms of universality, the most significant physically is the characteristic
00:50:03.920 universality of people, namely that they are universal explainers, which makes them universal
00:50:11.120 constructors as well. The effects of that universality are, as I have explained, explicable,
00:50:18.160 only by means the full gamut of fundamental explanations. It is also the only kind of universality
00:50:25.440 capable of transcending its parochial origins. Universal computers cannot really be universal,
00:50:32.800 unless there are people present to provide energy and maintenance indefinitely. And the same is
00:50:39.200 true of all those other technologies. Even life on earth will eventually be extinguished unless
00:50:44.000 people decide otherwise. Only people can rely on themselves into the unbounded future.
00:50:51.200 And that's the end terminology there. Let's just read the terminology. The jump to universality
00:50:57.200 is the tendency of gradually improving systems to undergo a sudden or large increase in the
00:51:03.280 functionality, becoming universal in some domain. Okay, yes. So generally progress happens incrementally
00:51:10.800 and this is what's happening in AI at the moment. So AI is having this incremental
00:51:15.680 improvement improvement improvement. But the jump to universality isn't just, it's not expected by
00:51:22.480 that, that it's not expected to just happen simply by virtue of the fact that you've had a
00:51:28.400 large number of incremental improvements. What we need for that jump to universality from AI to AGI
00:51:34.640 is a philosophical incremental improvement. And that's the mistake that people working in the
00:51:39.760 area seem to not have quite cottoned onto. Now David also mentions there that all of this,
00:51:48.400 this human universality is explicable only by means of the full gamut of fundamental explanations.
00:51:53.360 And where do you get those fundamental explanations? The fabric of reality, which has just had the
00:51:59.280 audio book recorded, which I would encourage everyone to get a hold of. The fabric of reality of
00:52:04.960 course goes through our most fundamental explanations of reality. And that's what's just mentioned
00:52:11.520 there. So if you really want to understand the nature of a person, you should have some understanding
00:52:16.640 of the theory of computation, the theory of evolution, quantum theory, and the theory of knowledge
00:52:25.520 or epistemology. This was a remarkable chapter and it's really leading us straight into chapter seven
00:52:32.640 about artificial creativity, which is going to be a very exciting chapter, all about
00:52:38.480 artificial general intelligence. So I look forward to seeing you again then. Bye bye.