Strong Artificial Intelligence: Representations of the Post-Human in Ibm’s Watson Computer Essay
A popular topic among post-human conversation is that of artificial intelligence. Artificial intelligence is a complex and controversial subject that has received, and is still receiving much attention among scholars. The general premise of artificial intelligence is to simulate or surpass one of the core components of human beings, intelligence. Intelligence is a defining feature that sets human beings apart from other living entities, our ability to use logic and reason far surpasses any other creatures’ in the animal kingdom.
Many scientist have experimented with the boundaries of intelligence, and it could be broke down into many sub-divisions. For the purpose of this essay I will be exploring notions as well as representations of what is known as ‘Strong Artificial intelligence’. Strong artificial intelligence refers specifically to the artificial intelligence that is meant to exceed human intelligence and is associated with certain characteristics such as consciousness, self-awareness, sentience, and sapience (Steels 75-110). There is a multitude of possible representations of artificial intelligence in the world of science fiction.
However, I don’t have huge repertoire of references for science fictions because my personal preferences don’t really tend to that genre. So instead I chose to analyse a representation that I am a little more familiar with, the IBM Watson computer. Watson is a complex artificial intelligence computer system capable of answering virtually any question it is presented. In this essay I would like to explore the philosophical aspects of strong artificial intelligence while addressing the spiritual dimension, inherent values, and controversial aspects of my chosen representation.
In 1997, IBM developed an advance computer system named ‘Deep Blue’ that was able to defeat the world’s best chess player, Garry Kasparov (Sostek). I feel it was at around this time that the general population begun to look at computers systems differently. One question that has stuck in my head since the beginning of this class is, ‘Is it possible for an intelligent being, like humans, to create a species more intelligent that its own? ’ For IBM, this seems to be one of their main goals. Although Deep Blue was a breakthrough technology that made lots of headlines, IBM produced a system that wasn’t very commercially valuable.
The technological accomplishment – playing chess really well didn’t translate into a lot of profit for IBM. In the mid-nineties IBM took up a new project with high hopes of changing the way we interact with data. How can they do this? Watson is the world’s most advanced deep-question answering system. What this means is that a question can be posed in natural language (everyday human elocution), interpreted (having ‘read’ a whole bunch of information, data and documents), and answered back (precisely and factually) in natural language.
So in other words, Watson is capable of understanding your question, and will deliver to you what you want in a natural flowing dialog. One thing that I find interesting is that humans rely very heavily on natural language to communicate, and that seems to be a field where computer systems struggle dramatically. David Ferrucci, the lead developer of Watson claims that this piece of technology could be regarded as the ‘Holy Grail’ of computers because of its seamless ability to converse naturally with human beings, letting us ask it questions instead of typing in keywords (Gunning, Chaudhri, and Welty 11-12).
I would definitely say that using this type of system is more reliable and efficient than using a Google or Yahoo search engine because instead of pointing you a document that might contain the answer you are looking for, Watson extracts the answer for you. Another thing that I find particularly important is that Watson is not connected to the internet; it is filled with legitimate data and documentations. What this means is that not only are the answer reliable, Watson is able to distinguish what it does know from what it doesn’t know (Gunning, Chaudhri, and Welty 11-12).
The computer only attempts to answer the question if it is ‘confident’ enough. When it was first suggest that Watson compete on an episode of Jeopardy! The idea was considered ridiculous, or even impossible. Deep Blues victory in 1997 was faced with a very different type of challenge – chess is a game which is based on logic, with fairly simple rules, that can be translated into mathematics, which a computer should be able to handle easily. Why would Jeopardy! be a lot more complex of a game for a computer based system to understand and play?
What makes language so difficult for computers is that it is composed of connotative and denotative meaning. In other words it is packed with ‘intended language’ that human beings are able to decode sharply and quickly, through shared experiences (Gunning, Chaudhri, and Welty 11-12). Categories, as well as clues on Jeopardy! often utilizes clever wordplays, allusions and connotations. Even if a computer was able to determine what the underlying question was, its methods of extracting data would inferior and slower than that of a human being (Sostek).
The advantage is given to a competitor who is able to make split-decisions and stitch together obscure pieces of information the fastest. Watson is an artificial intelligence agent that at basic level is composed multiple processors, nearly a hundred servers, a room full of technical gadgetry and has its information sourced from millions of pages from encyclopaedias, thesauruses, dictionaries, books, articles and also uses databases, taxonomies and ontology’s (Sostek). Watson could sift through all of this information and extract what it is looking for in a matter of milliseconds.
What is more, Watson’s system utilizes Natural language processing programs, reason, and knowledge representation programs to recognize and understand human language, even its most obscure references (Gunning, Chaudhri, and Welty 11-12). In the year 2011, IBM’s Watson was deemed ready to compete on an internationally aired episode of Jeopardy! Jeopardy! made history when it aired its famous special edition episode of Man vs Machine Match Up on February 14th-16th. It was a two game, combined point match up where Watson went up against two of Jeopardy’s all time, best competitors. Brad Rutter was Jeopardy! op earning contestant with his winnings just short of 3. 4 million dollars, the other contestant was Ken Jennings who won an incredible 74 game winning streak. Watson had some serious competition, no doubt. But can even the smartest people compete with such a powerful computer like Watson? Watson wasn’t actually present on set while the show was being filmed, because it massive size it was being operated in a nearby server room while being represented by an avatar on screen. In the end Watson dominated both of the matches, leaving Jennings and Rutter with just a fraction of what it had won.
I feel as if the general public had expected this, we are living in an age that is producing increasingly more intelligent and powerful computers. At the end of both matches Watson had won a total of $112,881 compared to Rutter’s $32,000 and Jennings $28,000 (Sostek). Watson had certain advantages and disadvantages compared to the human contestants in this scenario. Although Watson uses some of the most intelligent software on earth it has certain deficiencies when understanding the context of the clue. Because its primary search system is a keyword search, Watson has trouble with clues that contain a small amount of words.
As a result, sometimes Jennings or Rutter had the chance to get to the buzzer first; because Watson system only let it respond if it was confident enough with the answers it came up with. Although Watson comes up with the correct answer most of the time, its systems sometimes misinterpreted the information and occasionally misses an answer, it can get confused. A famous example of this was in the second round, Final Jeopardy, the question category was: “U. S. Cities” (“Its largest airport was named for a World War II hero; its second largest, for a World War II battle”) (Gunning, Chaudhri, and Welty 11-12).
Both Jennings and Rutter gave the answer we were looking for, which was Chicago. Watson answered: “Toronto??????? ” This seems like a pretty bad answer considering that Toronto isn’t even a U. S city. So, what could have happened here? Ferrucci offered some possible explanations for the error observing that the category (U. S Cities) did not appear in the clue, and that Watson may not have considered it, or that he may have retrieved “Toronto” because Toronto has an American League baseball team (Toronto Blue Jays), or that because there are cities named Toronto in the United States (Gunning, Chaudhri, and Welty 11-12).
Eric Nyberg, a professor at Carnegie Mellon University and also a member of Watson’s development team, stated that “the error occurred because Watson does not possess the comparative knowledge to discard that potential response as not viable. Although not displayed to the audience as with non-Final Jeopardy! questions, Watson’s second choice was Chicago. Both Toronto and Chicago were well below Watson’s confidence threshold, at 14% and 11% respectively. (This lack of confidence was the reason for the multiple question marks in Watson’s response. ” (Gunning, Chaudhri, and Welty 11-12) Watson’s failure was a disappointment to the IBM Company, but fuelled them to continue on the quest for knowledge, and to build even smarter, better artificial intelligent agents. Is infinite knowledge even possible? And at what point does knowledge cross into a state of consciousness? Was Watson aware that he was competing against other opponents? In turn, does that mean that Watson is aware of itself? These are some of the questions I will be exploring shortly, as well as how Watson reflects on society’s value system. To me, Watson is a modern reflection of collective intelligence.
What I mean by this is that Watson uses piece of information that comes from everywhere and everyone, in order to evolve towards higher order of complexity and thought, problem-solving and integration through collaboration and innovation. In society we are seeing constantly seeing new technologies being released to the market claiming to save us time, or make our lives more convenient. Some widely know examples of this include: GPS systems, social networking sites, search engines, even self-parking technologies. I certainly believe that Watson could fall under this category.
Because Watson is a deep question answering system, he would be a good fit for the workplace (***). Currently Watson is being tested in the medical field, assisting doctors making difficult diagnosis. Watson’s ‘job’ is not to replace a doctor, but rather provide a second opinion (***). Other practical uses I could see for Watson are; legal firms, where it could quickly sift through case details and legal citations, or a help-desk worker because Watson could answer promptly, instead of manually searching databases for product information, to give to the agitated costumer on the other end of the line (***).
Watson could possibly even be used to game the stock market, although there may be some ethical issues with that type of function. Earlier I asked the questions: Did Watson know that he was competing against opponents? And is Watson aware of itself? Questions of consciousness have always revolved around the concept of strong artificial intelligence. At what point is an intelligent computer able to think for itself? Have we already reached that point? In the year 1950 Alan Turing published a paper called Computing Machinery and Intelligence which opened with the sentence “I propose to consider the question, ‘Can machines think? (Turing 433-460). In this paper he proposed what would become widely known as the Turing Test. The Turing test is designed to test a machine’s ability to exhibit intelligent behavior that is equivalent to or indistinguishable from, that of an actual human being (Turing 433-460). The original model proposed by Turing consisted of an interrogator that would placed alone in a room, then engage (through typing) in a conversation (in natural language) with another human being (X) and a computer designed to generate a human performance (Y).
The interrogator asks both entities a series of questions, and if the computer is able to fool the interrogator into believing it is a fellow human being, it passes the test. The computer must be able to fool the interrogator at least 30% of the time in order to officially pass a Turing Test. According to Turing if the computer did pass the test, it would indicate that it is working at a human level, suggesting that computer can ‘think’ for itself (Turing 433-460).
Today there are many existing versions of the Turing Test, and it has been both a widely influential and heavily criticized component of philosophy. Although the Turing Test has been said to not be able to reliably test whether a machine can ‘think’ or not, it is a simple and appealing test, that even if imperfect, provides us with something that can be measured. It is important to note that The Turing Test does not check the computer’s ability to give the correct answer but instead, its ability to give an answer that is similar to the one a human being would give (Steels 75-110).
Watson would not however pass a Turing Test. The interrogator would realize that they are communicating with a super intelligent entity – not a human. John Searle’s 1980 paper Minds, Brains, and Programs proposed the “Chinese room” thought experiment and argued that the Turing test could not be used to determine if a machine can think. Searle noted that computer software could pass the Turing Test simply by manipulating symbols of which they had no understanding.
Without understanding, they could not be described as ‘thinking’ in the same sense people do. Therefore—Searle concludes—the Turing Test cannot prove that a machine can think (Searle 1-19). Searle’s hypothesis started off with a scenario; an English speaking man sits alone in a room with a book of rules. Then a page with a paragraph of Chinese characters is slipped under the door. The book then instructs him to manipulate and copy the Chinese symbols onto a new page, forming a reply to the earlier statement written in Chinese and the dialog continues.
The man does not need to understand Chinese in order to continue the conversation. Searle then argues that the same logic can be applied between computers and natural language. He then argues that not only are computers incapable of understanding language, they cannot understand anything, at least not in the same way humans use understanding. “In the literal sense the programmed computer understands what the car and the adding machine understand, namely, exactly nothing” (Searle 1-19).
In other words, to Searle, though a computer may pass the Turing Test it still does not literally understand the meaning of the words, and without understanding there is no way we can say the machine is thinking or that it is conscious. Searle’s Chinese Room argument has also been supported but also widely criticized. Ray Kurzweil takes an opposing view, saying that a computer passing the Turing Test would be tremendously important. Kurzweil, and Turing shared the same belief that human language embodies all of human intelligence.
In other words, there are no simple language tricks that would enable a computer to pass a well-designed Turing test. A computer would need to actually master human levels of understanding to pass this threshold. According to Kurzweil, “if it is really a properly designed test it would mean that this Artificial Intelligence is truly operating at human levels. And I for one would then regard it as human. I’m expecting this to happen within two decades, but I also expect that when it does, observers will continue to find things wrong with it” (Kurzweil 76-96).
Kurzweil predicts that a computer will beat the Turing Test by 2029 (Kurzweil 76-96). In 1999 Kurzweil wrote the book The Age of Spiritual Machines, in which he proposed that that “The Law of Accelerating Returns” (Kurzweil 51-65). According to the Law of Accelerating, the rate of change, in a wide variety of evolutionary systems (including technological ones) tend to increase drastically (Kurzweil 76-96). In other words, we are developing new and more complex technologies at a quicker rate, and we show no signs of slowing down.
Boundaries will continue to be pushed, and technology will keep reaching new levels of what seems to be ‘humanity’. To answer my earlier questions, I do not believe that Watson has any sort of self-awareness or consciousness, and it most certainly cannot exercise free will. Therefore Watson doesn’t qualify as true strong artificial intelligence, but it is the closest thing that currently exists. Watson exceeded its developer’s expectations when it defeated the Jeopardy! Champions and made an important step in the direction of strong artificial intelligence.
The possibility of a computer soon passing a Turing Test is becoming increasingly realistic. My personal opinion is that a computer system that is able to pass the Turing Test does not necessarily mean that it posses the same kind of conscious qualities as a human being does, but it will be a significant milestone in the history of artificial intelligence. For me, when a computer system is able to learn and take on intelligence for itself, produce its own algorithms, and make its own decisions would indicate that computers have truly reached a human operating level.
I agree with Kurzweil’s prediction that a computer will pass a Turing Test by the year 2029, but I believe the quest for better, more innovative, and more intelligent technology will continue on. Currently IBM is developing a new, more powerful and faster computer system which they are calling project ‘Blue Gene’. I actually could not find very much information on what Blue Gene actually does. One article I read said that it was an eco-friendly technology that was able to actually simulate 4. 5% of the human brain.
A Wikipedia search told me that “the project had two main goals: to advance our understanding of the mechanisms behind protein folding via large-scale simulation, and to explore novel ideas in massively parallel machine architecture and software” (Wikipedia). This brings me back to an earlier question; are human beings building a species that is more intelligent than our own? If a technology is eventually able to simulate the defining features of human being (consciousness, self-awareness, sentience and sapience) would we not then regard it as human? For some reason I find the idea of this more frightening than appealing.
Steels, Luc. “The Artificial Life Roots of Artificial Intelligence.” Trans. Array Artificial Life. . 2nd ed MIT Press Journals, 1994.
Gunning, David, Vinay K. Chaudhri, and Chris Welty. “Introduction to Special Issues on Question Answering.” AI Magazine. 2010: 11-12.
Turing, Alan M. “Computing, Machinery and Intelligence.” 1950: 433-460.
Searle, John R. “Minds, Brains, And Programs .” University of California. Berkeley, CA : 1980. Kurzweil, Ray. “Of Mind and Machines: Philosopoical Mind Experiments.” Trans. Array The Age of Spiritual Machine. Print.
Sostek, Anya. “Human champs of ‘Jeopardy!’ vs. Watson the IBM computer: A Close Match.” Pittsburg Post-Gazette. post-gazette.com, 09 2012. Web. <http://www.post-gazette.com/stories/business/technology/human-champs-of-jeopardy-vs-watson-the-ibm-computer-a-close-match-284466/>.
Wikipedia contributors. “Watson (computer).” Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 18 Nov. 2012. Web. 25 Nov. 2012.
Wikipedia contributors. “Blue Gene.” Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 4 Dec. 2012. Web. 1 Dec. 2012.