"Fungus Eaters" and Artificial Intelligence: a tribute to Masanao Toda

In 1962 Masanao Toda published a revolutionary study that was the origin of the idea of virtual complete autonomous agents and that preluded branches very important in Artificial Intelligence like autonomous robots (Brooks, 1991), swarm intelligence (Bonabeau, Dorigo and Theraulaz, 1999) or evolutionary robotics (Nolfi and Floreano, 2001). Against a cognitive science centered in a symbolic perspective, Toda analyses systems capable of behaving autonomously in an environment without external aids. This is a very hard problem in Artificial Intelligence. Complete autonomous systems have to incorporate capabilities for deciding what to do interacting with the environment.

In his article "The design of a fungus eater" (Behavioral Science 7, 164-83), Toda "consider(s) a design of a robot to be sent to a hypothetical planet as a robot uranium miner, which sustains itself by eating fungi as its energy source". This solitary creature feeds on a type of fungus that grows on the planet and the more uranium ore collects, the more reward it will get (see also Pfeifer, 1996, 3-12). The "Solitary Fungus Eater" (SFE) has means of collection and locomotion and means for decision making interacting with the perception of the environment. It must be autonomous and self-sufficient because it can not receive external aids: it is a virtual adaptive system, a genial idea in 1961-62. Besides Toda introduces another insight: the notion of bodily designed agent. He employs a choice program for regulating the behavior of the "Eater" using stochastic distributions. Toda defines the expected amount of uranium exploitation by means of a function f indicating the expected amount of uranium exploitation as a function of fungus storage (Toda, 1982, p. 114). The problem is that f cannot be determined until the choice program is fixed and the optimal choice program is given only after the function f is specified. The key is in defining adaptability in an unknown world.

But Toda advances almost 30 years and proposes for investigating matters like emotion, irrationality, social psychology, etc., for autonomous agents. Does SFE have emotion? Of course, the answer depends upon the definition and perhaps it can be defined as a particular state of mind accompanied by a high level of energy mobilization. After Rosalin Picard will introduce the paradigm of "affective computing" and nowadays many researchers design virtual agents or robots incorporating emotions or something like that: Lola Cañamero, Eva Hudlicka, Darryl Davis, Rolf Pfeifer, Thomas Wehrle and many others (see the excellent introduction by Ruebenstrunk, 1998, http://www.ruebenstrunk.de/)

Besides Toda thinks that apparently irrational behavior of human being can also be incorporated in the SFE. Herbert Simon had spoke about bounded rationality, notion exploited after by Stuart Russell or Gerd Gigerenzer and, in the meantime, the omnipresent contribution by Toda.

And what to say about the communication and cooperation between agents? Toda (1982, p. 127) believes that "optimal cooperation programs for Fungus Eaters will provide a heuristic for the sciences of human interaction" and...for robotics? Chains of e-pucks robots colaborating between them (see Francesco Mondada in http://www.e-puck.org/) demonstrate this prophecy in 2008.

Masanao Toda passed away on 5th September 2006. Who was Masanao Toda? Was he the prophet of a new world inhabited by artificial autonomous agents?