The Trouble With Turing

A second conceptual flaw indicates that Moravec's far-reaching predictions may never eventuate. AI culture arose in an intellectual environment dominated by Behaviourism and Claude Shannon's Information Theory, and was conceptually moulded by logician Alan Turing's seminal paper Computing Machinery and Intelligence (1950) that explored the question "Can computes think?"

Sidestepping the present, Turing focused on the potential future, formulating a hypothesis emphasising Behaviourist-like externally viewable psychological states that simulate or imitate inner psychological vistas. This model influenced British psychologist W. Grey Walter, who created mobile robotic turtles that seemed to exhibit complex social behavior. Walter's studies were, in some ways, the influential precursors of Moravec's current research.

However in an important 1990 Scientific American debate, John Searle skeptically confronted Turing's paper, demolishing its philosophical framework. Moravec's uploading scenario relies upon similar assumptions. Max Dublin remarks: "Here, of course, we have Turing revisited, but this time we have the imitation game played in spades. Even supposing, for the sake of argument, that one of the clever surgeon robots could be constructed, how long will the person on the table be given to decide whether the copy, the imitation of himself, which is being constructed in an electronically-fitted box beside him, really is the same as the original, that is, really the same as himself? Will we be asked to settle for a snapshot of reality in this case, too? Will five minutes do for this imitation game?"

Challenging the Universality of Universal Robots

Turing's Behaviourist model has been challenged by dynamic learning/complexity-modelling systems theories like Polish-American psychologist Mihaly Csikzentmihalyi's Flow (optimal experience) and Clare W. Graves' Levels of Psychological Existence. This critical gap undermines the technological advances required for Moravec's human-level intelligence robots to become common within fifty years. Consequently, both the Universal Robot conditioning software modelled upon mechanistic pleasure/pain principles and the world simulator that initiates learning are potentially flawed in making accurate predictions or implementing self-motivated learning strategies. Moravec also fails to detail, beyond a brief outline, an evolutionary developmental process for higher cortical functions such as reasoning, conscience, and intuition. These factors suggest that AI and humans may evolve along divergent paths, not necessarily a convergent symbiosis.

By extrapolating current computer processing power trends that suggest AI robots will mature into complex ecosystems, Moravec's scenario trajectory assumes critical breakthrough advances will occur whilst conveniently sidestepping the many failures and problems that have plagued AI's funded history. He fails to consider complex systems dynamics or link the evolution of AI machines and the Earth's non-renewable natural resources. What if because of insurmountable cognitive intelligence/learning problems we don't reach the critical mass required for self-evolving AI intelligent species, and geo-political instabilities prevent stop-gap solutions such as Moon or asteroid mining?

The AI Entrepreneurs

These conceptual errors are symptomatic of AI's scientific culture. Cybernetics, robotics and AI have been imbued with mythopoeic potential since their very beginnings. R&D have also been tied to sociopolitical expedience: when John McCarthy coined Artificial Intelligence in 1956, it was a successful marketing ploy to attract Pentagon funding.

Green ecologist Tom Athanasiou, author of Divided Planet: The Ecology of Rich and Poor (Athens, GA: University of Georgia Press, 1997), believes that AI entrepreneurs like McCarthy and Moravec are "unusually savvy in the positioning of their new enterprise. They looked far into the future, and refused to accept the limitations of their times. Indeed, their impatience helped spur the development of more powerful and interactive computer systems. They were ambitious and, from the very beginning, were married to the military research agencies that had birthed computer technology - and whose budgets could alone support their research. Unfortunately, the AI founders flashiest projects consistently failed."

For Marxist critic Les Levidow, the links between mythopoeic potential and sociopolitical expedience go far deeper, characterising computer culture's compulsive technology. In the anthology Cyborg Worlds: Towards A Military Information Society (London: Free Press, 1989), he proposes that AI and robotics mythopoeic scenarios indicate the military imprinting of society modelled after low-intensity psychological operations and technocratic command structures.

Levidow presupposes that early patronage by the Defence Advanced Research Projects Agency (DARPA), the Office of Naval Research (ONR) and Human Resources Research Office (HumRRO) have influenced military-information structures and those seeking to control the human element as part of larger hybrid organisms. These perceptions extends from cockpit cognition and computer managed instruction, to the rise of general education learning strategies seeking to artificially amplify and accelerate a student's cognitive processes, and business sales manipulation tools like Neuro-Linguistic Programming (NLP) and the Silva Method.

The mythopoeic power of Moravec's evolutionary paradigm creates a social space for hi-tech startups to flourish in, galvanising priority government and industry investment. In Cyborg Worlds, Tom Athanasiou links these powerful consensus-hallucinations with military R&D politics. "John McCarthy's early move to replace terms like cognitive simulation with the more flashy and fundable artificial intelligence may well be his most crucial contribution to computer science," he suggests.