Webelieve that by the time of the workshopDeep Blue will havelost another matchto Garry Kasparov, showinglittle improvementover the previous one. But even if it is indeed a "newkind of intelligence", it canbe arguedthat this intelligence is very basic. Thereis a long wayto go until DeepBlue could be considered a fully autonomousagent, the type weare eagerly trying to build in AIand that webelieve wewill eventually be successfulat. Weshould not consider DeepBlue (or any software agent) fully autonomousuntil it can manage its owncomputational resources (memoryand time), minimizerisk andcost of variousdecisions, assess its errors anddevelopnewrepresentations of (chess) knowledge,and be able to cooperate and communicatewith other computer and humanchess players. Wefeel indeed that DeepBlue falls somewhere non-trivial on the scale of intelligence. But to movefurther along that scale greater autonomy will be required: Is DeepBlue nowselecting its ownopenings, deciding whenand whet...