Final Report of the Workshop
Can a Machine be Conscious
The Banbury Center, Cold Spring Harbor Laboratory
May 13-16 2001
Given the great diversity of different views points promulgated at the meeting, tGiven the great diversity of different viewpoints promulgated at the meeting, no single one predominated. The emphasis range from the purely philosophical to the formal, computer-science and algorithmically oriented one, to the more engineering oriented neural network or robotics on, to the fundamental biological or the more systems, psychological view point.
Some of the speakers expressed the belief that the only way to ever build conscious machines is to first understand human or, more broadly, biological based consciousness.
Some speakers argued that the "Global Workspace Model" of Bernie Baars constitutes a viable model for a conscious system.
There appeared to be a broad consensus among the workshop participants that any conscious machine or artifact would need to have a sense of self and purpose and an ability to reason about itself (second-order consciousness). That is not to say that second-order consciousness about self can occur in the absence of first-order consciousness (to be conscious of a percept). But it was generally held that once machines can reason about themselves, can introspect and interact in a meaningful manner with humans, the basic conditions for machine consciousness would be meet.
Other participants argued that none of such functionalities, in and by themselves, would give rise to subjective feelings.
The only (near) universal consensus at the workshop was that, in principle, one day computers or robots could be consciousness. In other words, that we know of no fundamental law or principle operating in this universe that forbids the existence of subjective feelings in artifacts designed or evolved by humans.
In the opinion of one of the organizers (C.K.) progress will come on two unrelated fronts. A better understanding of the neuronal correlates of consciousness in humans and animals will be a conditio sine qua non: We need to know what animals show consciousness (e.g. bees, squid, rodents, monkeys, apes), under what conditions, when does it occur in humans, and, most importantly, what are the minimal neuronal mechanisms necessary for consciousness. Answering these questions will also help us to better understand the function of consciousness. Progress towards machine consciousness will be conditioned on both better understanding the preconditions for biological consciousness as well as on robust means to interact in a natural and robust ways with humans (that is, the development of sophisticated speech and language understanding capability as well as truly autonomous and flexible robots). Christof Koch