Week 1 - Emergent Wiki

Course Navigation

Week 1 (September 3, 5)

Timothy Ambrogi

The thing that struck me most about "The Thinking Machine" was how distrusting people of the time were of computers and the research of scientists. In essence, the video is nothing more than a series of rebuttals to arguments that the average person of the time might make against AI. I suspect that if a documentary were made today about the state of AI, it wouldn't need to provide as much evidence or as many visual hints as in the 1960's. Today people seem more inclined to accept the word of scientists as fact, rather than argue that it's "not true until I've seen it with my own eyes."

Also, I loved the first shot of the computer, where the camera actually *pans* to show you the whole rig. It definately cemented the stories one always hears about "computers the size of houses". It also explains why computers and robots in film are always covered in blinking lights and switches. It seems that they were right about a future involving a digital revolution, though not necessarily the tendency to microcomputers. (If I recall correctly, though, one of them did mention the idea of a more versatile computer that could perform all manner of tasks...)

--Tim A.

P.S. Did they actually shoot the questions and answer at seperate periods of time? And were viewers of the era supposed to believe that it was unscripted (since that's the way the style suggests)?

Lawrence Bomback

I really enjoyed the documentary, "The Thinking Machine," that we watched on Thursday in class. As a historical document alone it's packed with great sociological information, but that would be for another course. In terms of computer science, it was really wild to see what the computer experts 40 years ago were saying about their field. A few of the people interviewed are probably still alive today, and I'd love to hear the reactions to their predictions from 1961.

There were so many interesting points brought up in the video that still are actually quite relevant in today's scary world. One professor in the video said how robot's are both man's dreams and nightmares. That statement couldn't be more true. Currently, there are robots that can detect explosives and others that are the first "respondents" to certain terrorists attacks in order to prevent more human casualities in the rescue mission. But as Prof. Kumar said, what is stopping the terrorists to teach a robot to blow itself up? The computer scientist who was being interviewed said we really have nothing to fear because "we can always pull the plug," but I don't know how true that statement really is. We can "pull the plug" on our individual creations but not something that someone else creates.

My favorite quote from the film was from the man who said that computers will change the world in a way not seen since the industrial revolution. And, another expert in the documentary said that he believed the digital computer would be more important than the A-bomb. Both predictions were right on the ball. As for the first, what isn't run by computers nowadays? It has become as much of a necessity in the American household as a television, if not moreso. In terms of the A-bomb reference, many officials in the military fear computers as much as a nuclear device. They feel that it is only a matter of time before cyberterrorism becomes a powerful and popular weapon of choice.

Jacqueline Chew

Catherine Chiu

Although people may consider that there have been several advances in the field of artificial intelligence, the question still remains of whether computers will really be able to think and act like humans. This question was addressed several times in the movie, texts, and discussion, but the question can only be answered as a matter of opinion, not based on facts. While we may seem closer to the idea than forty years ago, it seems like we are still far away from creating computers that can think like humans. The human mind baffles us every day and presently, the thought of machinery being used to replicate the human mind seems like a substantial and distant goal. Because the human brain is such a mystery to us, it is hard to believe that we might have a machine that can mimic our behaviors and thoughts. The idea of artificial intelligence is also very dangerous. If the human mind is trying to duplicate itself, then that duplicate could potentially make copies of itself and that concept has an affect that is hard to foresee and control.

Jason Coleman

The thing I found most interesting was the notion that it seems that AI research is going no where because every time something is accomplished, some problem is solved, that problem no longer seems significant. And so, there's this idea that the bar is constantly being raised. The men in the film specifically referred to how certain intelligent behaviors that have been emulated on computers aren't considered to be intelligent anymore... behaviors like formula proving and letter recognition... Western script writing for instance.

I also saw a much closer relationship between computer scientists and pychologists (and other various scientists studying the mind) than I thought I would. It seems that AI research is an area of computer science that requires a lot of interaction with other scientific communities.

Nicholas Kerr

I thought it was really funny to see two fine gentlemen in suits sit in armchairs and ponder the capacity that artificial intelligence would take on in their near future. I think that both entertainment and intellectualism have come a long way since the making of the film. We all giggled each time the MIT professor lit his pipe, or when we heard their outdated, politically incorrect language. But what struck me as really interesting was the way in which the viewer was expected to believe all of the unsubstantiated suppositions presented by the men. As college students evaluating their hypotheses, we don’t criticize them too much for not perfectly predicting the future. But I can still try to imagine what the reaction must have been by viewers in the day in which the film was made. If I watched a modern movie in which I was told that computers had a very good chance of becoming smarter than humans, I probably wouldn’t believe it unless I was presented with more concreted evidence than a noisy computer learning to recognize letters or write a play, even if those were relatively new breakthroughs. There is a big difference between filling in the blanks to create a standard form play and challenging the intellectual capacity of me, Nicholas Kerr, or any other human. I certainly wouldn’t consider the suits and heavy tobacco use any form of credentials that could vouch for the statements of the gents. Yet, in my naļve knowledge of the middle of the twentieth century, I get the impression that people must have assumed that the predictions of the scientists were correct; that computers surely would become smarter than humans. I know vision is 20-20 in hindsight only, but I wouldn’t trust no foo’ cause he wore a suit, even if he did work for MIT.

Ananya Misra

The second point is the statement that computers do things that would be called thinking if done by a human. I think this is true (if a computer said that, we'd probably say it was programmed to do so). Even simple operations like arithmetic may fall in this category of tasks. So could the computer's play writing strategy--if a human had followed the same method, we might agree that she'd thought the play through using her knowledge of the world. In fact, if I'd had no idea how the computer wrote the play, I would have been more impressed by it.

So I think mysteriousness may be an informal part of AI. This relates to the machine intelligence aspect or pushing forward of frontiers mentioned in some of our readings: once a problem is solved it holds very little mystery and is no longer popularly considered AI. If a system functioned in a way we could not quite understand (despite having designed it), its products might seem new and we might accept it as an intelligent system. Which is ironic, considering that the 1955 proposal for the Dartmouth Summer Research Project on Artificial Intelligence was based on "the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it."

James Racanelli

Juan Ramos

I found the film to be very interesting in the sense that when looking at it a few decades into the future, a lot of the things that would have been considered as intellingent computers then are not treated as such today. For example, the film showed how a computer could scan a letter traced in a pen, and attempt to decide what letter was input. Today, we have Palm Pilots that can quickly scan and interpret human handwriting. Back then, people would probably be astonished at Deep Blue and declare it to be a true thinking machine, while today we can understand it in terms of highly complex and adaptive algorithms that might or might not be considered as producing intelligence. Because of this, I tend to agree with Jason in that we keep raising the bar in the field of AI. However, I also think that extensive research in AI has yielded some other fields in Computer Science.

After watching the show, I think I can redefine my own interpretation of what defines an intelligent computer. Given all the rules and procedures that makes a computer work, an intelligent computer would be, in my opinion, one that could either bend, ignore, or break the rules it was programmed with in order to accomplish any given task. This, perhaps, will require another few advancements within AI and the broader field of Computer Science. Maybe we will need to create computers that can store information at the level of the human mind, which we don't even fully understand yet. Nevertheless, it is clear that we will still make new and 'more intelligent' changes and advancements in the coming years, both in AI and other Computing fields.

Matthew Rushton

After watching the film and doing the reading it became clear to me that AI can be viewed as two entirely different studies. The first side of AI, and certainly the more understood, involves getting computers to "appear" to think. By this I mean the computer itself possess no intelligence but can be programmed to do tasks like play chess or write westerns. Doing such can at times appear to give the machine intelligence, when in fact there is no actual thought going on. For example Deep Blue couldn't play Tic-Tac-Toe even though it is a far simpler game. A generic game playing machine (I do believe they actually have these)would be a step closer to true AI. Yet even this system wouldn't understand how to write or read a sentence, and further couldn't be taught unless it was reprogrammed. I believe that this is where Turing's test for intelligence fails. The ultimate goal of AI in my opinion shouldn't involve perception, or what a machine "appears" to do. Rather it should be about the actual processes the machine uses. This is much the same as the subsymbolic approach that is described in the text, the second side of AI. Here the actual processes are of the utmost concern. If the goal of AI is to produce a truly thinking machine (and it obviously is), then the question becomes what is the best means to go about it? Does programming simple stand-alone systems capable of doing one task bring us closer to our goal? Maybe, maybe not. That isn't to say programs like Deep Blue are useless, far from it, but are they bringing us any closer to our goal? For now the two sides of AI remain far, far apart. I suppose only time will tell if they converge or diverge yet further.

Tina Tan

The film "The Thinking Machine" was interesting to watch since it gave very different perspective of artificial intelligence than the one that most people of our time hold. Our generation has become so used to computers as a way of life and it is clear how dependent on computers we have become. Before I saw the film, it was hard to imagine how some people, even those experts consulted in the film, couldn't imagine how everyday people would want to have these machines. Seeing those computers that hold what we now consider to be a very small amount of information yet take up such amounts of room space, really shows how far technology has advanced. It was also interesting that in terms of understanding human function and creating an intelligent machine, we are still far away from that goal. In creating an independently thinking machine, the same fears stated by the scientists still remain, especially with the dangers that concern us today. There is the question that whether humans should continue to develop machines that may someday pose threats or sacrifice potential advances that help us understand human intelligence better. In response to other students' comments, I agree that the term AI is relative. The fact that AI research will always yield advancements causes what is considered AI to constantly change.

Ian Harrison

It was interesting to see how impressed the 1960’s television host was by a computer that could write bad, dialogue-less two character westerns using what was probably a strict set of rules and some random numbers, yet he didn’t have nearly the same reaction to Samuel’s checker program, even though that program was one of the first successful examples of machine learning. This odd bias is also reflected in Shapiro’s paper. Though he speaks glowingly of learning as being an AI-complete task, he doesn’t seem concerned that the solution to any other AI complete task might not be so complete without learning. Being able to understand natural language would be quite a feat, but as far as artificial intelligence goes, being able to learn from natural language is vastly more important that simply understanding it. As Shapiro wrote, all human knowledge can be encoded in natural language. In seeking to mimic human intelligence, AI must be able to mimic human learning, as that learning is at the basis of a persons intelligent functions. People are not born intelligent, they are born with a predisposition towards intelligence. It is the ability to learn, and especially the ability to learn language and the symbolic manipulation it entails, that makes a person intelligent. Even a person’s own nervous system seems to go through a learning period during early life, when they learn to control their hands precisely, to walk, and to vocalize accurately. Reinforcement learning such as the kind exhibited by Samuel’s checkers program, which learned to play only by playing against itself and knowing that winning was good and losing was bad, seems the closest of any AI approach to the way in which people actually learn during early stages of life.