More Science Talk
Steve: Welcome to Science Talk, the weekly podcast of Scientific Americanfor the seven days starting January 30th, 2008. I'm Steve Mirsky. This week: An anniversary celebration with Carl Raggio formerly of the Jet Propulsion Laboratory and some funny stuff with a couple of researchers teaching computers to recognize jokes. Why teach a computer to get jokes? To get to the other side of the interface. Plus, we'll test your knowledge about some recent in the news. Lawrence Mazlack is the director of the Applied Artificial Intelligence Laboratory at the University of Cincinnati. Julia Taylor is the graduate student who has been trying to develop ways for computers to understand humor at a very basic level of wordplay and knock-knock jokes. I called them at the lab in Cincinnati.
Steve: Professor Mazlack, Ms. Taylor—good to talk to you today.
Mazlack: Hi Steve.
Taylor: Hello Steve.
Steve: You have research going on, that one's really serious business, but you're trying to teach computers to get jokes and right away that's a kind of a funny idea. But there is obviously a serious reason to do it. Why do you do that?
Mazlack: Well, we were interested in sociable computing, which means that we like computers to be able to naturally interact with human beings without a lot of problems. There are a lot of formal walls, and the best way to do this—at least from the human's point of view—is using normal language or natural language;
and but it probably needs to be informal language, which means that it can't be precise step by step by step; and so understanding humor is part of understanding informal language.
Steve: It's interesting when you do a Word—a spell check on a Word document, maybe that's one of the simplest examples of the kind of problems that come up in communication.
The word, For example, I constantly type "from" instead of "form" and spell check doesn't catch that. It thinks it's spelled correctly and everything is okay, but the sentence doesn't make any sense. So is it a similar kind of concept to try to improve informal communication to get people to understand each other, and that's where humor comes into play—or to get computers and people to understand each other, I should say?
Mazlack: Well, humor is one example of informal communication, but the issue is to try and remove the barrier between human beings and machines. Your example of "from" and "form" is a good example, but it's a little bit more elaborate to try and do informal communication.
Taylor: However, the basis of it is the same. The reason why "form" and "from" are confused is because the software that you use will not take context into account, and to understand humor you have to deal with context.
Steve: So right away you have to educate the program with some context before it can even start to appreciate wordplay.
Taylor: Right. To understand any kind of humor you have to have knowledge about the world—
of which would come in the context of some sort. So a computer program would have to have that knowledge.
Steve: So what kind of context? I know for example that the earliest attempts at communicating humor to the program was knock-knock jokes; so what kind of context do you supply so that the knock-knock jokes get appreciated by the program?
Taylor: We put a lot of world knowledge and anthology, so we are taking everything that we can out of a children's dictionary, because a lot of knock-knock jokes are for children, so we are hoping that the information that is there, a lot of information that is there will be useful for us. So when we are working with an algorithm for a knock-knock joke; we will take a third line and then come up with some words that sound alike and then we would go into the anthology to see how they fit with the sentence that's in the fifth line.
Steve: The third line being the response to "Who's there?" and the fifth line being the punch line of the knock-knock joke. You're currently trying to move past the knock-knock jokes into some more sophisticated wordplay. What's that about?
Mazlack: Well, essentially sort of standard wordplay jokes; for example, my favorite is a kid's joke—what is black and white and red all over, turns out to be
(unclear 4:59)[a newspaper].Now Julia also has one where she deals with mice.
Taylor: "Computer teacher: Why are you bringing cheese into the computer room"; and the little boy responds, "You told me I was going to work with a mouse today."
Steve: Right, right.
Taylor: So, these are the types of jokes that we are trying to analyze, those that have some context behind them; but the joke is say, some wordplay.
Steve: Let's go back to the black and white and red. Because I know a number of punch lines for that, what's the punch line that you are using?
Taylor: A newspaper.
Steve: A newspaper, okay. How do you know that the program or the computer that's running the program has appreciated the joke?
Taylor: I don't think we want the program to appreciate the joke at this point. We just want it to know whether or not it's a joke. We don't want the computer to have emotions necessarily; we want [the] computer to know when we have them.
Mazlack: Well, I think we are different here. Now what Julia is trying to do is to get the computer to recognize when a joke has occurred. Now to me when you said appreciate, that means whether it is a funny joke or somewhat funny joke, on that order; myself, I'm personally interested in doing that, and perhaps my next student will be interested in recognizing gradations of humor.
Steve: Okay. So, right now the computer does not type back at you, "Ha ha?"
Taylor: Right now, it just says "yes, it's a joke"; or it just types back "yes it's a joke", [or] "no it's not a joke".
Steve: Right. It just understands when the form is that of a joke.
Steve: People are probably familiar with this kind of concept [of] teaching machines humor; and the example that comes to mind is numerous attempts on Star Trek—Next Generation with Mr. Data, trying to get him to better experience humanity by being able to understand humor. Is that just an interesting thing to you or does that actually in some way inform what you do?
Taylor: Well, I'm not sure if we are trying for a machine to understand humor; I would say we are trying to recognize when it occurred, that's probably more of what we are doing. In that sense, Data probably recognized when something was a joke, whether or not he understood it, yes, or had emotions; so that's a different question.
Mazlack: Yeah, because Data's problem was that he didn't have emotion.
Mazlack: And so, intellectually, he could understand. For example, I believe Data liked chamber music and was an accomplished player, but whether or not he had emotional response to what he did—that's a different question. That's one of the older questions going back to Turney's article on artificial intelligence—whether or not something needs to have an emotional response in order to be artificially intelligent. We don't know if a computer would have emotional response; yeah, if it was successful in recognizing it might, I don't know.
Steve: What's your take on that Ms. Taylor?
Taylor: I don't think I want computer[s] to have emotions, but that opens the discussions I guess—whether or not they should.
Steve: After reading a lot of Antonio Damasio, I tend to think that without a physical body that really experiences the environment, it would probably be pretty dangerous to have a disembodied intelligence that did experience emotion.
Mazlack: Well, embodiment is one of the basic questions on whether or not a machine can comprehend the world the same way that we do. The people at M.I.T. are doing some work in embodiment and George Lakoff—the linguist at Berkeley—he believes that we perceived the world through our embodiment and our reaction to the world; you know, being 6 foot or 4 foot tall, whatever it is, and the way we interact with the world. It might be that embodiment is necessary, I don't know.
Steve: I want to tell you one of my favorite jokes.
Steve: And maybe it can be useful eventually because its got a couple of different things going on and maybe it could be a good test for the program eventually. And the joke is: What's the difference between Noah's Ark and Joan of Arc?
Taylor: What's the difference?
Steve: Noah's Ark was made of wood and Joan of Arc was Maid of Orleans.
Mazlack: (laughs) But there is a similar joke: Two vultures get on the airplane and each one of them is carrying two dead raccoons. A stewardess comes up to them and says, "Gentlemen you're only allowed one carrion".
Steve: Right, right, right. Well Dr. Mazlack, Ms. Taylor, very interesting. What do you think, where would you, I mean, this is the hackneyed journalists' question, but where do you see things five or 10 years from now in this field?
Mazlack: My belief is that we'll be able to understand at least wordplay jokes and this will lead to more sociable computers, making machines more accessible to people who really are afraid to push the button at all.
Steve: So, elderly people, people who are just not that comfortable with technology maybe could have a more seamless relationship with their technology?
Mazlack: Right. It doesn't only have to be elderly people, it could be anybody who is—for whatever reason—doesn't think precisely and exactly the way your computer needs. Another aspect of what I work on, I work on soft computing
in[and] fuzzy logic, which again tries to put gradations of belief into things.
Steve: Gradations of belief into things?
Mazlack: Right. For example, if we say a restaurant, we are trying by—the quality of food at the restaurant as the computers currently basically deal with the food is bad or good: 0-1. However, we'll like to be able to say it's very good, it's sort of good, it's [a] little bit rotten or it's awful; and
this people think in terms of this way, but it's difficult to communicate with the machine in terms of this gradation of belief.
Steve: Right, right. So ultimately you want to be able to say, "Waiter, what['s] this fly doing in my soup?"
Mazlack: Yeah, right.
Steve: Very good. Prof[essor] Mazlack, Ms. Taylor—thanks very much, I appreciate it.
Mazlack: Okay. Thank you.
Taylor: Thank you.
Steve: Now it is time to play TOTALL……. Y BOGUS. Here are four science stories; only three are true. See if you know which story is TOTALL……. Y BOGUS.
Story number 1: Snakes hear through their jaws.
Story number 2: In April, the Texas higher education coordinating board will decide whether to approve a program that will grant Masters Degrees in science from a creationist perspective.
Story number 3: Researchers are working on contact lenses with informational displays so you can see like the Terminator.
And story number 4: College students are responsible for almost half of all Internet movie piracy.
We'll be back with [the] answer[s], but first, January 31st marks an important anniversary in the history of the final frontier. Fifty years ago on that date, the U.S. got its first satellite into orbit. It not only joined Sputnik, but actually returned some valuable scientific data regarding the presence [of the] Van Allen belt—the charged particles around the earth that are held in place by the magnetic field. Carl Raggio was intimately involved in getting the satellite up. I called him last week at his home in Glendale, California.
Steve: Mr. Raggio, how are you today?
Raggio: I'm very fine Steve.
Steve: What were [you] doing 50 years ago today?
Raggio: Well 50 years ago today, we were getting [ready] for launch; that would be the launch of Explorer 1, although we didn't call it Explorer 1, we called it "Deal."
Raggio: Yes, just like in cards, because that's what we played while we were ready for launch
Raggio: Usually gin rummy.
Steve: Gin rummy, okay. So you launched finally on January 31st; but what was the environment like, you know, if we were there, 50 years ago—we are speaking on January 23rd—what was going on a week before launch?
Raggio: We were
a pretty anxious and for many reasons. One is that we were under the gun to go ahead and get something up because of Sputnik, and in fact there were two Sputniks then, that only exacerbated the whole timing and that was that; we were all waiting and anticipating a good launch, and as it turned out, it did launch well.
Steve: And what was your actual job?
Raggio: Mine was a designer on Explorer 1.
Steve: And what were your responsibilities?
Raggio: One, to do the design work on the satellite, and to put the Van Allen experiment into a six inch cylinder.
Steve: So talk about that kind of technical challenge.
Raggio: We had several. And one happened to be the weight and strength. So we used fiberglass as the principle material in there, because it has a great weight-to-strength ratio. And we didn't have much in the weight of power supply other than batteries, so we used batteries—they were dry cells—for communication. The antennas turned out to be a whip antenna; and if you look at the Explorer, you'll find that it has four pieces of wire, and this was done for two reasons; one is not only as an antenna, but to maintain stability in space because the last three stages were in a spinning tub, so you wanted to affect [a] gyro type of thing. If you take a look at the outside of the Explorer, which happens to be the upper part of the fourth stage; and you'll find that it is striped, and the striping on it, (laughs) it was all done to maintain a stable environment for the batteries and communication systems; and that's because it was spinning—it would give us a gray tone, so we wouldn't absorb too much in the way of the Sun energy—as we were on the sunny side—and correspondingly we wouldn't lose too much in the way of heat on the backside.
Steve: So you figured, you'll split the difference and striped that thing.
Raggio: Yeah! (laughs) We knew how to spell thermodynamics but we didn't know a great deal more than that at that time.
Steve: Okay. And you were specifically looking for the Van Allen belts, which you figured had to be there.
Raggio: Yes. And the instrument was one that, that would give us a measurement; and incidentally it's a little known fact but Explorer 3, which was launched a couple of months later about four months later, confirmed the presence of the Van Allen belt.
Steve: So what exactly, did Explorer 1 confirm then?
Raggio: That there was cosmic radiation emanating from the belt.
Steve: Okay—as opposed to the presence of the belts themselves.
Steve: So, you worked with Wernher von Braun.
Raggio: He was part of it. He was the director of the Army Ballistics Missile Agency, located in Huntsville. They took several trips out here and the first stage was a modified Redstone launching missile; and it was an artillery weapon, in fact it was a derivative of V2s that were deployed in World War II. And the last three stages were JPL—the configuration was called Jupiter C.And the last three stages reproduced, and they were composed of, as I mentioned, a spinning tub with a 11 scaled down Sargent missiles which were six inches in diameter on the perimeter and then inside were three scaled-down Sargent missiles; and Jupiter C was a program to test materials for reentry. In other words, the ICBMs and IRBMs, the Intercontinental Ballistic Missile needed materials so they could reenter the atmosphere without burning up, and that's what we were doing. And we took away the cone which was the test vehicle; therefore that inserted the fourth stage and that became a deployable one.
Steve: Right, because with a three-stage rocket you can do the ICBMs, and you knew that if you just hack on a fourth stage you could reach orbit.
've proposed this about two years before; keep in mind, we were under the Department of Defense at that time and we had proposed this to the Department of Defense. They said no, it was not something that they wanted because I think things were so volatile between Russia and ourselves and it was shortly after the incident with Gary Powers and the U2 that was shot down over Russia. So consequently they didn't want one more thing to agitate it, since we were using a military booster. But they began a program at that time coincident with what we were doing called Vanguard; and Vanguard didn't have a success story. So when Sputnik was launched, they reverted to what we had and told us to get something up.
Steve: Right. Once Sputnik was up in the air, all the cards were on the table, if you will.
Raggio: Yeah, because the Russians had used an ICBM for a launch vehicle.
Steve: Now I just wanted to ask you about von Braun, because I saw you were quoted in a newspaper article as saying that "he could sell ice to Eskimos."
Raggio: He could. He had a personality and a charm that went beyond anything; he was great spokesman at that time for the space program
Steve: And its probably one of the reasons why he was able to survive in every environment that he found himself in.
Raggio: I think so. I think this man knew that in one way or another we were going to go to space.
Steve: On the 31st we have the 50th anniversary. Let's also go back now 50 years. What was it like when the thing goes up and you start to receive the telemetry and it's in orbit? What was the atmosphere like? Pardon the punk, as there is no atmosphere up there. (laughs) What was the atmosphere like in the room where you guys were—and you were a bunch of young guys; you were all in [your] mid-twenties-to-early-thirties?
Raggio: That's right. I was in my late twenties; in fact I'll be 80 this coming July, so I was one of the younger ones. We were about as happy as one can ever get. My wife never knew what I was working on, and so it wasn't until 11 o'clock that night, that I called her and said "We did it"; and it was kind of, "Did what?" I said, "We put one up there, and so she found out what all those long hours and everything else were, where I was at the labs doing this, because we were in the auspices of the Army and everything we did was secret; so we weren't able to tell anybody about the things which we were doing. But we were turned on, and I've been turned on ever since. Even though I retired in 1990, and I had about 40 years at the laboratory, space is still one of the more exciting things for me.
Steve: Well Mr. Raggio, this was a lot of fun; and congratulations on the anniversary.
Raggio: The one thing that I might add to this is at that time, because things were so tense between the two countries, we were allowed a degree of freedom that we really don't enjoy today; and that is that most Congress people really don't understand what research and development is. And the development aspect of it happens to be the assessment of your failures. And so consequently you learn from making mistakes you did. The whole process is learning; and what happens nowadays is that, generally, Congress only wants to hear a success story. So consequently you are limited to do less. Space business is risky business.
Steve: Things that look like failures are actually part of the experimental process.
Steve: But this is such a public experiment that it doesn't get interpreted the same way by the public.
Raggio: No, and I don't think that, again the general populous really appreciate what a risk it is every time you do something in space.
Steve: By the way, we are nowhere near finished with studying the earth's radiation belts. In 2012 we are launching the radiation belt storm probes to try to better understand the relationship between solar activity and the fast moving electrons and ions in space because that kind of space weather can affect the many important satellites currently spinning around the earth.
Now its time to see which story was TOTALL……. Y BOGUS. Let's review the four stories.
Story number 1: Snakes hear through their jaws.
Story number 2: Possible Masters Degrees in Creation Science in Texas.
Story number 3: Contact lenses that display info.
And story number 4: College students responsible for almost half of all Internet movie piracy.
Time is up.
Story number 1 is true. Snakes don't have external ear parts, but they do have fully formed inner ears that are connected to the jawbone. They rest the jaws on the ground and pick up vibrations that way. For more, check out the January 24th episode of the daily SciAm podcast, 60-Second Science.
Story number 2 is true. The Institute for Creation Research was already given preliminary approval by the Texas board to offer Masters Degrees in Creation Science in that state, but the institute asked for the final ruling to be postponed after it received some further questions from the board. That's according to the Chronicle of Higher Education. The institute's own Web site lists tenets including that humans were created in fully human form from the start; that the first humans were Adam and Eve and that the earth was created in six days. So since everything is already known what would the thesis be about?
And story number 3 is true. Contact lenses with displays could be coming. University of Washington engineers use microscopic manufacturing techniques to combine flexible contact lens material with imprinted electronic circuits and lights. They presented their results January 17th at the Conference of the Institute of Electrical and Electronics Engineers.
All of which means that story number 4, about college students being behind almost half of all Internet movie piracy, is TOTALL……. Y BOGUS. But what is true is that the Motion Picture Association of America has been making that claim for more than two years, according to The New York Times. Last week, the MPAA said it had gotten its math wrong and that college students are actually behind only 15 percent of movie piracy, but the association still wants federal legislation requiring colleges themselves to stop piracy.
Well that's it for this edition of the weekly SciAm podcast. You can write to us at podcast@SciAm.com and check out numerous features at www.SciAm.com, including the latest science news, videos, and blogs. For Science Talk, the weekly podcast of Scientific American, I'm Steve Mirsky. Thanks for clicking on us.