Smart, Friendly and Personal
An Agent for All of Us
Computer science is the alchemy of the modern age. This is what makes the field so attractive to inquisitive young minds. And agent technology has become one of the hottest areas of computer science, in response to a number of factors (but predominately business needs). Computer scientists, today's alchemists, don't produce gold but rather programs that "live" in the digital habitat.
"The Fetishism of Commodities", a key concept behind "alienation" in Marxism philosophy, may never change in future society. Marx said that "price, taken by itself, is nothing but the monetary expression of value. The values of all commodities of this country, for example, are expressed in gold prices, while on the Continent they are mainly expressed in silver prices." (Karl Marx, Capital). In the 1970s, the international monetary system lost its convertibility with gold. Since then, our paper money has been guaranteed by nothing more than mutual agreement. In the 1990s, money is losing even more of its materialistic roots. "Electronic money" is taking the place of conventional money. Values are expressed in digital data; manipulating digital data via computers has become integral to the basis of value formation. The old saw "Time is Money" needs an upgrade to "Processing Speed is Money"; the processing speed of computers becomes the most essential part of value formation.
Agent technology will develop at an accelerated pace for next decade or so. We have a need for agents to process tasks in our daily life. We'll use agents to check schedules, answer the phone, receive and send e-mail, gather information on our interests, monitor stock markets, secure our homes, and so on. In other words, we're going to depend upon agents to be our secretaries, our messengers, and our guards. So we need these agents to be smart. They'll need to grow in intellect in response to our needs.
Agents are developed "bottom-up", in contrast to the conventional "top-down" approach of designing Artificial Intelligences. Learning is a very important process to agents. Artificial Intelligence applied to business uses so far leaves us with an impression of inflexibility. They show intellect in given tasks but are not so smart in newly-experienced situations. Most Artificial Intelligence programs are built upon the expert system, which seeks to store the knowledge of medical doctors, lawyers, engineers and so on. AIs have been used in power plants, hospitals, transportation systems, etc. Pursuant accidents have made us aware that expert systems aren't always trustworthy. (This is a sci-fi standby, and we see it again in "Jurassic Park", where so much trouble begins with a computer system under-counting the park's dinosaurs. The story plays on our worries about computer systems.) Too many experts know their expertise is flawless. They're wrong.
Already agent programs are at work on the desktop and on networks. Each agent has unique tasks and behaviors. On desktops we see operating system agents and application agents; on the Internet, Web search agents, Web server agents, information filtering agents, information retrieval agents, notification agents, service agents, and mobile agents; and on intranets, collaborative customization agents, process automation agents, database agents, and resource brokering agents.
"MExA = Mind Extension Agent" - Cheerful communication between mind-havers and mindless machines
When I address you, I include us both in the class of mind-havers.
Daniel C. Dennett, "Kinds of Minds"
An agent I am currently developing is called MExA (=Mind Extension Agent). MExA is created for use by man the "mind-haver". It is a speech-recognizing, brainwave-reading, speech-enabled, information-searching and learning agent. (Speech recognition is currently limited to English.) MExA's current job is to entertain users culturally. It selects programs and information from the Internet (or an intranet) for its users, guided by man-machine interaction utilizing brainwaves and voice. It talks, tells jokes and selects Web pages containing information that it deems appropriate for the user.
But I really expect MExA to help bridge the gap between man and machine. The computers we use every day are not so human-friendly; they certainly don't seem to be making friends with us. We don't include them in our "class of mind-havers" yet. (Of course, the definition of "mind-haver" varies by individual.)
"We left Houston at dawn, headin' down the road - just me and my truck." Strange. If this fellow thinks his truck is such a worthy companion that it deserves shelter under the umbrella of "we," he must be very lonely. Either that, or his truck must have been customized in ways that would be the envy of roboticists everywhere. In contrast, "we - just me and my dog" doesn't startle us all, but "we - just me and my oyster" is hard to take seriously. In other words, we're pretty sure that dogs have minds, and we're dubious that oysters do.
Daniel Dennett, "Kinds of Minds"
As Dennett says, it's odd how the truck driver treats his mindless vehicle as a worthy companion on the road. We give minds to dogs but not oysters. "Some think it's obvious that a ten-week-old fetus has a mind, and others think it's obvious that it doesn't." (Dennett) So what about computers? Some surely think of their computer as a collaborative colleague, yet most consider it just a machine. We're nowhere near creating a computer like HAL in "A Space Odyssey". Marvin Minsky says of "HAL's Legacy": "I am still a realist: If we work really hard-and smart we can have something like a HAL in between four and four hundred years. I suppose if we're lucky, then, we can make it by 2001!"
MExA is no HAL. It has no intelligence, and only responds to a user's voice and brainwaves. (It's worth noting that even HAL described himself as an intellect-less mimic of man's behavior.) However, MExA can record the user's voice and brainwave patterns so that future versions could draw upon these data when communicating with man.
Why have MExA use brainwaves? EEGs are used as input data to investigate man's "mind". Other possible inputs include event-related brain potentials (ERP), electromyograms (EMG, a measure of muscle activity), pupilometry (a measure of changes of pupil size), electroculography (EOG, a measure of eye movement), electrodermal activity (EDA, changes in electrical activity at the skin surface), heart response, blood volume, and blood pressure. Then there are measures like respiration, oxygen consumption, salivation, skin temperature, and gastric motility. In addition to speech recognition, the use of face recognition should be considered in the future.
These body signal analyses and readings are inputs to be processed by the computer technology. These data are used in some fields such as Psychophysiology. But there are cultural reasons to use these too. The technological developments born of computer engineers aren't always keen on cultural awareness!
For example, the Unicode developed mainly by computer engineers in Western countries may not fit non-Western cultures if it restricts their use of language. People in China, Japan, Korea and Taiwan use Chinese characters (along with home-grown characters in Japan and Korea). The Taiwan government has formally recognized about 50,000 characters. Most Western countries use alphabets containing far fewer characters. Ignoring these differences, Unicode unwittingly tries to simplify real-world cultural traits in its zeal to simplify digital text processing.
Even the ways of using "yes" and "no" may differ among cultures. A native speaker of English asked "Don't you like apples?" might answer "No, I don't like apples". But in other langauges including Japanese, the answer would be "Yes, I don't like apples".
Existing computer technology is far from enabling linguistic communication between machine and man. For us to understand each other, computers are going to have to climb higher up the intelligence ladder. And when they gain those smarts, the use of physical input data might help machines understand their mysterious organic partners.
A technological archetype of MExA was originally developed for "News_agent(x)" (1997) exhibited at ISEA97 in Chicago. News_agent(x) was my first work in a series of agent-based interactive art works. The following text was written for ISEA97 to describe the basic concept of News_agent(x):
News_agent(x)
Stock markets, wars, power struggles, environmental problems, murders, and more -- we always try to find out what's behind the headlines on television news or what's to be read between the lines of newspaper articles. It's easy to see that there are many causal relationships lying under the surface of the daily news. All the small particles of cause interact with each other, to become social phenomena at the final stage.
Evening news programs on TV can work toward family ties, providing a time for sharing common subjects with other family members. Yet 24-hour news programs like CNN or BBC Satellite can turn us into world news addicts. We can hardly stop watching news programs. They bring dramatic spectacles to our homes in real time, as do music and sports channels.
News seems to influence our feelings, thoughts, behavior and attitudes in everyday life. All of human activity reflects the mechanical, materialistic elements which form the world. This is the traditional sociological viewpoint, but it is being turned upside-down. We should now start interpreting society through the methods of biology. Social structure needs to be seen as a biological system rather than a material system. Applying this interpretation to news, the reflection of our social structure in all of our activities is based on biological reasons, since individuals' activities can be understood in the context of biological rules. The latest theories in financial analysis - as well as other fields - derive largely from biological findings. (Theories based on "complex systems" are a good example.) Darwinian competition, too, is being applied to those theories.
For ISEA97, I would like to set up a closed circuit information flow, a virtually-created "news program", which consists of TV images, newspaper articles and financial market reports. A "favorability" score indicated by the audience's brainwaves will categorize news contents. Brainwave data, in this case, depicts the audience's subconscious. In between the human response and news factors, there would be an agent which indicates the audience's state of mind. There would be 4 degree of "favorability" in responding to news: very favorable, favorable, neutral, and unfavorable. Depending upon the "favorability" indicated by measured brainwaves, the agent would modify the news and financial market reports that it calls up. The more the audience indicates favor, the more the agent "grows", like a biological entity. If the agent grows to a certain degree, it starts breeding. The agent has a life span, so that it will have to mature within a given period of time. Otherwise it cannot bear descendants.
The result of the whole process above can be output as a graph in a financial report. The audience can see various types of news and find out how their minds relate the news factors at the same time a computer detects how an audience's brain is behaving.
News_agent(x) worked only locally as part of an installation artwork at the ISEA97 exhibition site. As an installation artwork - although an unfinished one - it represented my concerns about man and machines, information, and culture.
MExA (=Mind Extension Agent) carries on all the functions of News_agent(x), is able to retrieve information from networks, and has the other functions mentioned above. At the current stage of development, it doesn't differ from the ordinary software agents you see on your desktop or on some Internet sites. However, it's still developing. It will be learning as an intelligent agent program and will communicate with other agents too.News_agent(x) is an interface agent custom-selecting news by corresponding with users' brainwaves. By contrast, MExA is an entertainment agent -- with an additional focus on promoting communication between man the mind-haver and mindless machines. In the future, studies using MExA may cross-develop with other scientific and cultural fields such as Psychophysiology, Biology, Psychology, Ecology, Anthropology, Philosophy and Medical studies.