The example of the text-book-reading robot illustrates why we have more than enough computing power. If only we knew the "algorithm" or set of rules to write a program which would teach the machine to read, it would have an astonishing SHAI capability. But that should not cause us to denigrate what enormous computing power might do for AM. COMBINED WITH EVEN AN INFERIOR ALGORITHM, ENORMOUS COMPUTING POWER MAY GIVE US AN SHAI. Here, for example is a 4th. kind of SHAI which may develop, relying on a "brute force algorithm". Suppose we have all the questions which all of the people in the world can ask in plain and everyday language, ie the vernacular, and all of the answers, also in the vernacular. If John Doe in Singapore has symptoms a,b,c and d, he can ask for a diagnosis and get an expert answer. If Jane Doe in Africa wants the trend line for all stocks on the NYSE projected from the last year to the next year, she can get an answer. That answer also requires an expert for its calculation (eg the statistical expertise of calculating a best-fit line) but the answer is in the vernacular. Across all subjects such a machine would be indisputably beyond the intelligence of any human. And it could be programmed with a very simple program, ie accessing a giant look-up table. Give it a question in the vernacular on any subject and it looks up the answer.And there are other ways in which simple algorithms could be combined with enormous computing power to yield SHAIs. This prospect gives us the AM counterpart to the search for the Holy Grail. sometimes referred to as the search for the "magic bullet" for SHAI.
Is there a rule or algorithm hitherto overlooked, which would enable us to take advantage of the enormous computing speed and capacity which we expect to arrive soon and create a machine with SHAI capacity? To some it would be a Frankensteinian monster.
Frankenrobot or not we carry on. There is no stopping the "March of the Machines" as Professor Kevin Warwick titled his 1997 book. Chatterbot is our 5th. type of SHAI, ie the machine which can converse better than a human. It is like the Question-and-Answer Machine above in that its interface with humans is the natural language, but it carries on a more extensive kind of conversation, like a human in everyday dialogue with other humans. When you hear the expression "Turing Test", this is the kind of test which is envisaged: a machine in one room conversing with a human in another room. And the machine which passess the Turing Test is the one which puts out conversation indistinguishable from that of a human. In recent years others have noted that one could create a variety of tests like this. If Turing Test refers to the ability to meet the criterion of "human equivalency" (as coined by Professor Hans Moravec at CMU) with respect to conversation, then we can ask further, "What kind of conversation?" The conversation of a two year old? The conversation which is restricted to questions and answers on mental arithmetic? The conversation of a mentally retarded person?
If "Turing Test" is restricted in some of these ways, we can pass the Turing Test now. We could write a program to meet human equivalency for an average two year old in conversation; or an average person with an IQ of 50. As a matter of fact this is one way to deal with the NLP (natural language programming) problem in a more systematic way. We could start at age two or IQ 50 and work our way up slowly and systematically from there. This is sometimes called "developmental robotics".
When the expression NLP is used we are usually referring to the sort of general, everyday conversational ability with which everyone reading this page is familiar. The expert predictions for this kind of chatterbot which passes the traditional Turing Test are optimistic. Kurzweil (1999) foresees that "In the next decade we will see...computerized personal assistants that can converse ...." (p.4). And Jim Krane (2001) in the AP article cited earlier tells us Reid Simmons at CMU "...estimates robots will be smart enough in 25 to 50 years to hold intelligent conversations with people."
Predictions like this are rather amazing, given that conversational ability is a mainstay of what we humans consider to be that which places us above the other animals. Presently one can purchase "Sitepal" software, eg at <http://sitepal.com/vhost_minisite/index_banner.html> to give us text-to-speech ability with an "avatar" as it is sometimes called. This is a computer animated picture of a human or human-like figure. Sitepal gives us a talking, moving head-and-shoulders on the screen for our web site. What happens when that "software robot" is the President and can converse in natural language as well as any human president or better? Some time in this century, likely the first half, such NLP will be available the experts tell us. To discuss this further, refer to <http://www.yahoogroups.com/group/Robot-for-President>.
To get the full benefit of the Sitepal demonstration requires having speakers connected to your computer.
Digital multimedia is the new expression used to refer to the more or less integrated use of the sensory modalities for machine input and output. The two major sensory modalities as we have seen from the previous lessons on Artificial Sensation are sight and sound. Sitepal works with text in, sight and sound out in a more or less intgrated manner. "Less" is more apropos as the state of the art currently allows some movements of face and mouth for the avatar which are integrated or co-ordinated only in the sense that the avatar is moving from the beginning to the end of the speech. The expression VIVO is often used in reference to machines with some digital multimedia capacity, ie "Voice In-Voice Out". A VIVO-equipped machine will accept the voice in of the human user and return voice out by the machine. Sitepal accepts voice in via type, although it could conceivably be used also with software (readily available on the market) which accepts a human voice in and converts that into type. Equipped in this way, a system with Sitepal would be a VIVO system. You would talk to the online avatar and it would talk back to you. "Digital Mutimedia" is a new field. Chapman and Chapman write in their text "Digital Multimedia" (2002) that "We have become so accustomed to the high speed of technological change in computing that multimedia is already seen as an established feature of the computing landscape. But, on the time scale of cultural change, it has barely arrived." (p.8) See <http://www.wiley.com/digital_multimedia>
Thus we could take the five SHAIs summarized so far and either profile them for similarities and differences or describe them in lecture format as has been done here. These SHAIs are all counterparts to performing humans. Humans can be considered to be geniuses in very different ways. A Stephen Hawking (mathematician) is very different intellect expressed compared to a chess genius like Kasparov or a political genius like George Bush, Jr. But all will be surpassed some time in this century as the "march of the machines" proceeds for better or for worse. The result will be up to "us", ie some portion of homo sapiens
tucked away in this obscure corner of galaxy with its billions of stars against a backdrop of trillions of galaxies. Democratizing the say on where this goes is a major objective of IMP. If we the people are relegated to the status of technopeasants who leave the "march" in the hands of a few billionaires and military planners, ie the technolords and technopriests, we are doomed.