Intelligent agents may become all the rage again with Microsoft Corp’s recent investment in General Magic (CI No 3,361), but IBM Corp has already created its very own virtual electronic information economies into which it has unleashed numerous agent technologies to help anticipate rules and behaviors and compare them with human economies. Researchers at IBM’s recently- created Institute for Advanced Commerce – funded with an initial $10m and staffed with 50 scientists – are trying to figure out what’s going to happen when there really are a billion people connected to the net via tens of billions of devices, each commanding hundreds of intelligent software agents. Today’s agents, IBM says, are used to seek, sort, and select information on the internet. In the future world IBM describes, they will engage in complex transactions on behalf of their human owners – or on behalf of the devices on which they reside. As the scale of interaction across the internet increases, software agents, it believes, will be used more extensively to manage the flood of information and opportunities. IBM researchers anticipate billions of intelligent agents will roam the virtual world, handling all levels of simple-to-complex negotiations and transactions. They expect profit-driven agents that can make money by offering services such as translation or data mining will be the most prevalent. However although faster at making decisions than humans, agents can’t cope with unanticipated situations, and while they can sift – mine – huge quantities of information, IBM says market and system efficiencies – and flaws – will operate on an accelerated level.
Double agents
Its two key observations from using agents in these virtual economies are the emergence of specialized markets and roller coaster pricing. It’s found price wars result when agents are programmed to maximize profits by buying or selling at the best price. If provided access to information about bids, each selling agent sets its price just below the lowest offer. Taking humans out of the equation speeds the bidding process and prices drop precipitously in a matter of seconds. When profits from selling a smaller volume at a higher price exceed those from selling larger volumes at the prevailing low price, an agent raises its price and the other follow suit. In simulations, this cycle repeats itself endlessly. IBM says more complex agents will be able to search and learn from past transaction data, avoid those situations and develop strategies to profit from them. It thinks specialized online markets are likely to emerge where agents can go to trade in certain goods, information, and services. In addition it expects hierarchies of agents will emerge; manager- agents will oversee the work of simpler agents, and agents will form associations to cooperate and compete in the market. So who is going to create the first double agent?
