The Sentient Robot
by Rupert Robson
Imprint Academic £17.95
 

ANALYSIS of the future of Artificial Intelligence (AI) tends to focus on the potential of the technology and neglect its inherent risks. Rupert Robson’s approach is better balanced: refreshingly clear in outlining the moral implications and dangers of AI, his assessment of the barriers to achieving “Artificial Super-intelligence” (ASI) is based on detailed accounts of both machine learning and human cognition.
 
But some of the arguments presented in The Sentient Robot are superficial and dangerous.
 
The book acknowledges ASI could pose existential threats to humanity but argues that it is pointless to oppose a technology that “will happen anyway”. It is more productive, Robson proposes, to forge principles for the design of Human-Purposed ASI.
 
The inevitability of ASI is attributed to insatiable consumer demand for AI-based products and services, and the value of AIs to governments in their quest for geopolitical dominance. Unfortunately, the book does not address the possibility that transformation of our socioeconomic system could change the kind of tools that we imagine, make and employ.
 
Robson recognises ASI could affect the way we think and act, and even our physical existence. He quotes cognitive scientist David Chalmers, who suggests coping with the speed and capacity of non-biological systems may force us to “dispense with our biological core entirely”. Robson isn’t merely resigned to this drastic and frightening prospect; he believes it might offer a positive future for humanity.
 
The central focus of the book is the “last two hurdles” for ASI. The more easily tackled is human-level flexibility of thought, or Artificial General Intelligence (AGI). Current progress in relation to the application of neural networks and expert systems suggests this is achievable.

Robson’s explanations of human cognition and psychobiological systems are accessible, and his claims about the prospects for machine learning are compelling. He makes a powerful case for the possibility that ASI will emulate multiple forms of human intelligence.
 
The second and more challenging hurdle is the development of machine consciousness. Robson’s unearned conviction that this will be achieved is the book’s most serious flaw.

Given the clarity and detail with which he highlights our limited understanding of human consciousness, his confidence in relation to this is astonishing. In effect, he is urging us to embrace an experiment with unknown parameters and a vast range of possible outcomes.
 
The book’s strengths include its synthesis of evolutionary biology, cognitive psychology and philosophy. Its exploration of human decision-making, illustrated with specific examples, provides plausible explanations of effective and maladaptive behaviours; and the explanation of links between language, thought and consciousness is accessible without diminishing the complexity of the systems involved. And Robson should be applauded for tackling the issue of which systems should be built, rather than those which simply can be built.
 
But all this is undermined by neglect of the socio-political context in which systems are researched, designed and created. Prospects for human-purposed technology will always depend on its ownership and the way it is financed.