thumbnail of About science; About artificial intelligence
Transcript
Hide -
If this transcript has significant errors that should be corrected, let us know, so we can add it to FIX IT+
This is about science produced by the California Institute of Technology and originally broadcast by station KPCC in Pasadena California. The programs are made available to this station by a national educational radio. This program is about artificial intelligence with host Dr. Robert Mugabe and his guest Dr. Gilbert McCann professor of applied science and director of Cal Tech's computing center. Here now is Dr. McGrath. We define as intelligent thought. All of man's efforts to arrive ordered concepts or laws from observations of himself and the world around him by focusing attention successively on small segments of the total body of information and attempting to enlarge the field of his concepts by correlating the studies. In the context of our topic this evening we speak of course in terms of artificial intelligence and that translated into modern language private means computers. Could you tell us something about computers in
terms of how they developed and the historical background. Yes but the real practical development of computers started during World War Two. One of the first important computers was actually developed at MIT then of our bush. Well when was this about 1944 45. I just finished this computer was designed for a very specific task to solve the differential equations of describing the trajectories of ballistic missiles and the automatic control problems of gun control and fire control devices which was a technology of World War 2. The first real digital computer which deals with numbers and formal mathematics was built by Dr Howard Akan at Harvard University and other pioneering university that has pioneered in the development of computers. His mark to computer was in mechanical attempt to mechanize the desk calculator. The first truly
modern concept of a rather high speed large scale computer was developed also at the latter stage of the war by two professors at the University of Pennsylvania. MARTIN Ackley and they developed the first vacuum tube a true electronic computer called the ENIAC. Now out of this technology was fostered by let's say the war effort we had at the end of World War 2 in 1946 and industrial computer. Activity that amounted to little 7 or 8 million dollars a year in the last 20 years. This is burgeoning to about I 20 about a 7 billion dollar industry. Something for the growth plan with tremendous impact on all of our every day life. I noticed making your remarks that you indicated in the case of the second step in computer development the notion of a digital computer. This distinguishes that type of computer.
What other type than the analog computer is a common name for a class of other computers which sort of simulate by a mechanical or electrical principles of electricity whereas the digital computer works in terms of mathematical formulas. Now the problem you have to put real numbers into it in some form and it manipulates these numbers by the laws of arithmetic. Now these are the computers. Preponderance of news today. Yes by far of that seven billion dollars worth of industry 99 percent of it is probably in the form of large scale digital computers. Some of the analog computer still is an element of our need and trance but only very small not grown anywhere near the rate. The other asked how would you describe the functions of a computer. Even these that work in times of arithmetic concepts where we can think of this first in terms of what most of us have. For many we have a desk calculator
a desk calculator is an elementary form of a digital computer. You can put numbers into it. You can multiply or add two numbers and produce a product which is an arithmetic operation. But you have to do these Each step by step yourself. Now the modern digital computer first is has different than the desk they are in the fact that every one of these erect manic operations in a modern high speed computer can be performed in a few millions of a second. In addition to this modern high speed digital computer has a large memory for instance it may have 10 km in a memory made of little magnetic cores that can store tens of millions of numbers or words and it can get at these numbers bring them into its arithmetic processing function in a few millions of a second so that the two prime functions then are the business of data processing might be at the magic steps that the computer takes and
secondly its ability to store information and retrieve that information for further processing. Are many people Asian in one important part of his memory is that it's so organized that even the instructions that it may do in an automatic way many many thousand of these individual steps can be done automatically by what is known as a program which is also information words or instructions which are stored in these high speed memories. Talking about their automatic process does the computer work in terms of some kind of arithmetic that wing on any number system that we know are there other application. As overrun approaches I use Well mathematics has developed a large number of number systems for instance we think most people in terms of a decimal system 0 to 9 the characters of the numbers. Actually most computers are more efficient designed to work in what's called a binary system. Just a 0 and 1 just two numbers. But there are relationships between the binary system
and the decimal system and it's easy for a computer for instance to take information that a human may given decimal form convert it to this binary form produce arithmetic operations in binary form converted back to decimal form and using information in the system that you understand for instance. Is this a convenience because of the electrical causation of the system. Yes it's more efficient you can build faster computers smaller and cheaper if you use the binary system. You indicated this large growth and computer utilization went on in fact some of the applications and the areas of our social interest that utilize computers. You know Obama one of the biggest applications is our federal government and state governments and even then we know simple government. It's amazing how well it just would stagger the imagination to find out how many millions and millions of words of information vital statistics printed in print or
link all US citizens of the country are stored away and in the memories of computers that are possessed and operated by our various governments. So they probably are the largest user. So this is by and large the storage and processing the specific kinds of data. It's not so much a manipulation of information and athletically but merely a storage and library function. It's not the kind of well it's quite a bit of processing or planning ahead of various logistic operations there's a lot of what might be called really sophisticated mathematics that's necessary to play in the modern government operation. One of the second big users are becoming more and more important to our every day life for that large automated factories that are being developed. The automobile industry is becoming highly automated. They petroleum large chemical automatic factories which are having an impact on we humans because of the potential effect it will have on our
labor situation replacing let's say human labor by automated factories is a very important development and in the field of automation and computers it's interesting to look at some of the things that are being done in terms of the use of computers for our health. The computerized hospital is here a practical thing. What do you mean by computerized hospital. Well not a number of hospitals are now very successively developed successfully developed methods by which computers will keep track of the records of patients that are being treated in the hospital. The records of the drugs that have been. Given me determines who women there should give them new drugs so that there will be no let's say mistake made in the treatment of the patient. The laboratory tests that are made for instance urine tests and blood tests and X-ray tests. Nowadays these data are taken and can send a major in many hospitals to computers which analyzes these data and actually perform very
effective diagnoses themselves and tell a doctor what the disease is so we have a very nice observer here and consultant who manages perfect memory so to speak and has a large memory as a matter of fact so they make comparisons. Correlation is not the kind of person in this particular case. Since this process can be well-defined let's say by people who think about what the process should be a very high reliability let's say of these computers make some very reliable sort of a function. It across the country through Western Union or the telephone company. Two central computers in their decisions to make transaction in the New York Stock Market are controlled by these computers. Even two in terms of making analyses to determine the best time to buy or sell stocks for the most profitable type of operation. This means however that they must have some rules by which to instruct the company that's right on a lot of
their ethical research has been done to a client to back up let us say the proper mathematics for this type of a business it also implies that there's of understanding what the process is as well as to that there are a lot of things one can't predict about the trends of the stock market so that sometimes these computers don't make the right decision. Airlines have done a very good job in automating their operations. Many of us are probably impressed by the very great efficiency of the ticketing situation these days how easy it is to for you to obtain a reservation how reliable your reservation is kept for you and so on. The whole control of the ticketing and all of the operations including flight control of the large airlines are highly automated with large scale computers. Hans and Simon many a time standing upon a counter and asking if there's a flight available from so-and-so at such a time which I want to change to and in moments I can get an answer. It wasn't just a casual telephone conversation.
I was actually a sheep. The agent probably picked up a telephone and dialed a number and talked into something that was actually perhaps talking to either a person gathering to push them buttons to get some information out of a computer. In some cases however they have a direct connection to the computer I think and address the question to a computer as fact. How about other areas of our society. Well some of the more glamorous ones that we read about in the newspaper and so on are the military and our space research programs which have had a very profound impact on the fundamental development of new computer technology. The need for automatic control obviously in outer space where in situations where human beings can't cannot or not cannot even be affected perhaps if they could be there is a spurt of very profound amount of very fine basic research and automation and computers. Perhaps however some of the more interesting applications of computers today are
in education and in basic research and educational institutions such as ours at Cal Tech and such notable places as I mentioned earlier in my tea and Pennsylvania and Harvard. There is the business of research and computers as does the development of computers persay already utilization or is it both. Well actually there are some present technology of course of computers which we see breaking and just coming. The modern computer that I say has been developed largely through the application of modern electronic technology and the use of formal mathematics. These types of computers just as we known today and as they work let's say effectively are very useful in basic research for processing data of modern research. It's interesting to note that some of our modern type of research such as nuclear nuclear research or research by modern biology in the complex. Questions we wish to ask about.
Living nervous systems and the fine lines on the application I'd say the definition of a problem in mathematical terms and then asking the computer to perform the mathematics. Well let's say that modern research basic research is so complex and gather so much information that even let's say in the keeping track the storage and keeping track of the information so you don't lose it might recall require these large memory and then you have to ask the same examine the data correlate many different bits of data and there are many situations where let's say existing formal mathematics is quite adequate. The technology in other words in the mathematics that exists is quite adequate. However there is developing in a number of institutions in one Cal Tech is one of them. A basic research program which is focusing focusing its attention on who we are. Recall information science. Now this is research which is trying to get at say a better
understanding of how human beings or any living nervous system understands processes information does creative thinking rather than does formal things which are defined the same by mathematics in order to let's say determine how we might build the next generation of computers which could be even say more fascinating and what they can do and might be what we will call the products of automatic our artificial intelligence of the future. The language of the computer today is mathematics percent as we know it is it now and virtually all most applications I'm talking here about different kinds of mathematics are richer languages so to speak. Well there are many kinds of mathematics which deal purely with numbers and in terms of let's say a language which describes an they are not to complex each language is nowhere near as rich say as the languages by which for instance a conversational language you are using these languages are called algebraic languages.
They do are small units of numbers you make all the numbers by a symbol and the language is typically one of taking a given number multiplied by another number that it wants. There was one of these sorts of things. Now out of this kind of a rather simple language one can take arithmetic expander arithmetic and algebra where the geometry can one can approximate by finite arithmetic finite difference arithmetic calculus. One can deal with stochastic mathematics theory and probability and these are the more common formal mathematics that are used by let's say the present types of computers that are trying to do so you're talking really about going beyond as I would now were of course in our basic research that seeks a better understanding of artificial intelligence to find let's say a richer kind of language that can deal with more sophisticated concepts and we feel that the basic. Concept by which we can do this or here the biggest problem I
say is not so much one to one to building a computer that can do something more creative but understanding what creativity is. Defining it in terms of a richer language and one can build a computer that can perform more creative thought. What are the elements of creating content in this context of our present remarks then. What are the things that you would call creative as in effect a computer are what the computer can do. Well one good example Bob of this is the problem of vision. In fact this is one of the areas of research that we are engaged in to try to get a better understanding of what we call creative thought. I'm kind of curious that you selected the example of vision. There's a particular uniqueness or an advantage. Well for one purpose one advantage of it is that the visual the sensory system of man and the visual system. It starts with say the site sensory system of the I used the information let's say which is coming into the living nervous system the sight sensory system can be well-defined it's
known visual patterns they process I'd say of trying to probe into the visual nervous system is easier than going higher up in the nervous system. I see the biologist as a learned let's say how to get inside be emotional nervous system without just damaging a living organism to get some to actually see how it functions. Information on how it functions. So there are certain biological point of view and advantage that you can get I don't understand it and perhaps profit from that experience. Now it's also interesting that in for instance in humans the sight sensory system is the most sophisticated of all our sensory organs. So you must be talking about lower forms of life then when people are humans all we're interested in vision. But for instance 90 percent of all the information that comes in through the sensory organs of Man comes in through these eyes. He Jaya for instance has over 100 million rods and cones which can. In other words there are a hundred overwhelming little points of
detail of a visual pattern except that it's projected on the corner of an eye so like the dots on a television set. But there are 100 million less than a few thousand in it. Now it's interesting because of this type of us site's entry system the human system is so complex it's also important to do research on simple organisms and their research program at Cal TAC which is seeking to understand the principles by which a visual nervous system abstracts visual patterns is being done not only in human but only in sacks and many different successively more complex animals. I believe I've heard of what's going on with the PA system as they do their work at MIT on the frog because it has one has been discovered to have one eye sting property which has enhanced understanding of this process of abstracting visual patterns. They eye of the frog. It's been discovered that right behind these rods and cones where
the first like information this is trans formed and from the light energy into electrical signals which is messages with messages which are electrical which is the form of the physical form in which all information flows in the nervous system right behind these rising cones are several million little time of neurons and those neurons through this very fine dendritic connections to the rods and cones the information from many rods and cones produced new electrical signals which they send on up to another point in their system which is the first stage of data processing or abstracting a concept they take. Signals from a large number of these light sensing elements and organizing information and some particular way and send it on to other processing center they send it up into the brain of the of the frog. But but interesting enough this first stage of data processing or
computing abstracts certain important properties of the promenades for survival. For instance someone enters a kind of program online already right in the retina small computer. Think of the retina of a small initial computer that very rapidly computing certain things necessary let's say for the survival of the frog for instance. Some of these neurons recognize a fly when it's flying across tiny I think simply by noting the fact that there are small dark objects and they are moving in the direction in which they're moving determinist to the frog. The fact that this may be applied to flying in a certain direction is within his range and he sticks out his tongue and tries to grab it. Now that's the final signal that comes back out of the computer. We're going to signal that I will have the machine the frog act in a certain way. Is that the kind of thing I did on gas as a matter of fact. We think we know enough. This principle of how the recognition of small dark objects can not be thought of as insects
are very large dark dark objects that the fly might think of as a as a hawk and he should jump in the water instead of sticking out he's tough enough is known for instance from research of this sort on the principles of that type of pattern extraction that people have a number of people who actually build electronic pattern recognition devices which can function as well as the frogs dying so the element of this artificial intelligence when I will meant is this business of pattern recognition or organising organizing information recognition. Yeah this principle of how abstract important patterns concepts out of tremendous amounts of information to millions and millions of bits of visual information this is probably the same process by which the central cortex of the brain abstracts other kinds of concepts not in the army. Non-visual information writes hope that this is the case and therefore by learning how delusion works we will learn how the central cortex does creative thinking. One other important element makes up this
notion of artificial intelligence. Well and to actually understand it in a conceptually conceptual manner requires what we what we think of as a language in adequate language to description. In fact the science the information science is based upon the theory of a language of mathematical linguistics now the formal language of mathematics have their algebraic languages. Mathematics However today the formal types of mathematics that have been developed are extremely in terms of their capacity to understand something as complex as I say human nervous system are extremely simple. We think of mathematics as a very complex mathematics today. Mathematics of algebra and calculus stochastic mathematics is really not adequate to describe the very complex processes. Let's say I was raised on it and I was. Rich and I want it. It just is not rich enough in terms of the amount of information you can handle by the complexity
of the language rippling which is we think a richer language and by saying Let's say that from now on is describe an object but the amount of information that's contained in the real description of the object or a single number may be 10 digits. But the recognition of a person as Bob might be millions and millions of bits of information is required by the mind to scan through its memory to finally say yes that's Bob. So Bob is a richer much richer word and say a simple number and sort of sharing and this is what I'm having back to telling you now I hope. How is this interest and examining and gaining experience with him. I have. Lower animal forms lead us to developments in this computer technology both in terms of how to recognition and in terms of language development. Can you talk about that and uncle actually if it were only made a beginning and really
understanding the living nervous system. We've learned a certain number of very small number relatively thousands of bits of important information but that's only a very small percentage of the hundreds of millions of let's say of things we have to know before we truly understand a living or a system. Nevertheless this says this The research that has been done up to the present time is extremely important. It has laid the foundations for new strategies of research and theoretical strategies particularly that must be developed in order to create let's say these richer languages and there is in French interestingly enough mathematical linguistics has now been developed to the point where one can indeed approximately model some of the simpler aspects of the mind of man as he even thinks in a conversational language. One of the professors at Cal Tech has actually developed through the extension of the theory of logic in mathematical linguistics
a concept of a computer memory and a language structure whereby it really can think in English it is told simple information slowly in terms of English it builds up on knowledge just like a baby may perhaps learn to recognize the words. There's literally a learning. I came across and that indeed is simulating. In the first stages at least of a true mathematical concept of a rich conversational language and perhaps a way in which some of the basic features in which the mind either works is not really essential this business of the learning process is not part of the creative element. Well we certainly have something related to it or a corollary to that. Well it's been well established that that living nervous system of most particular higher order animal such as man is something which develops continually from birth. The living nervous system is continually channeling the way in which it functions as a result of sensory experiences and new information which comes into it and is continually learning how to think by
saying and function in terms a richer and richer languages which describe it's true here. I get the impression and at the moment we're very much in the beginning stages of this learning process and I see the development of computers and the sources of artificial intelligence. You would you say that we were I think crawling stand so as people were just beginning to really cross because I was making progress. And perhaps we can look over some of the not too distant future I hope. We can have devices which that can sound and which are languages in which the word accommodate and the hand of large quantities of information then extract other information from them. Well thank you very much. This was about science with host Dr. Robert McGregor and his guest Dr. Gilbert McCann a professor of applied science at the California Institute of Technology. Join us again for our next program when two more prominent
scientists will discuss a subject of interest about science is produced by the California Institute of Technology and is originally broadcast by station KPCC in Pasadena California. The programs are made available to this station by a national educational radio. This is the national educational radio network.
Series
About science
Episode
About artificial intelligence
Producing Organization
California Institute of Technology
KPPC
Contributing Organization
University of Maryland (College Park, Maryland)
AAPB ID
cpb-aacip/500-dn3zx700
If you have more information about this item than what is given here, or if you have concerns about this record, we want to know! Contact us, indicating the AAPB ID (cpb-aacip/500-dn3zx700).
Description
Episode Description
This program focuses on the science behind artificial intelligence. The guest for this program is Dr. Gilbert D. McCann, California Institute of Technology.
Series Description
Interview series on variety of science-related subjects, produced by the California Institute of Technology. Features three Cal Tech faculty members: Dr. Peter Lissaman, Dr. Albert R. Hibbs, and Dr. Robert Meghreblian.
Broadcast Date
1967-11-14
Topics
Science
Media type
Sound
Duration
00:28:46
Embed Code
Copy and paste this HTML to include AAPB content on your blog or webpage.
Credits
Guest: McCann, Gilbert D.
Host: Hibbs, Albert R.
Producing Organization: California Institute of Technology
Producing Organization: KPPC
AAPB Contributor Holdings
University of Maryland
Identifier: 66-40-62 (National Association of Educational Broadcasters)
Format: 1/4 inch audio tape
Duration: 00:28:29
If you have a copy of this asset and would like us to add it to our catalog, please contact us.
Citations
Chicago: “About science; About artificial intelligence,” 1967-11-14, University of Maryland, American Archive of Public Broadcasting (GBH and the Library of Congress), Boston, MA and Washington, DC, accessed March 19, 2024, http://americanarchive.org/catalog/cpb-aacip-500-dn3zx700.
MLA: “About science; About artificial intelligence.” 1967-11-14. University of Maryland, American Archive of Public Broadcasting (GBH and the Library of Congress), Boston, MA and Washington, DC. Web. March 19, 2024. <http://americanarchive.org/catalog/cpb-aacip-500-dn3zx700>.
APA: About science; About artificial intelligence. Boston, MA: University of Maryland, American Archive of Public Broadcasting (GBH and the Library of Congress), Boston, MA and Washington, DC. Retrieved from http://americanarchive.org/catalog/cpb-aacip-500-dn3zx700