“What is it like,” asked the philosopher Thomas Nagel, “to be a bat?”

Come to that, what is it like to be a laptop? Both bat and laptop behave rationally and both appear to have an interior life to which we have no direct access. Is there a real experience of being either?

The issue is not what it would feel like for you to be a bat/laptop, but what it feels like for a bat to be a bat. And even if you could miraculously become a bat for five minutes, could you report back on the sensation, given that it involved being a bat not a human? Probably not. We have, therefore, no access to the phenomenon of being a bat. Human knowledge, science, stalls in the face of the inner life. And, as our machines acquire their own inner life, our inadequacy will become ever more apparent.

Maybe not. Maybe there is no such thing as the feeling of being a bat. To have such a sensation, a bat would have to possess some degree of subjective awareness and we have no evidence that it does. A bat does not know what it is like to be a bat. This leaves only human self-awareness. We seem to know what it’s like to be human.

Or do we? After all, extreme behaviourists once tried to convince us that our sense of an inner life is just one more form of behaviour that is, therefore, entirely accessible to science. I’m fooling myself when I tell myself I know what it’s like to be me. These behaviourists were, you might think, bonkers. You’d be right.

This may all seem preposterously abstract. Our internal sense of the colour red or the smell of a flower (“qualia” as philosophers call these things) are just the way we live and nothing more needs to be said. We should just get on with it. Unfortunately, just getting on with it now requires us to consider these matters. Computers are making decisions and robots are on the way. Do they — will they — have minds? And, even if they did have minds, how would we know?

That is the subject of this, on the whole, lucid book. David McFarland is the man to get you through this thicket of thought. A zoologist, biological roboticist and psychologist, he has covered the waterfront when it comes to alien (meaning animal or machine, not Martian) minds. The book is an unusual blend of science and philosophy, each taking over when the other fails, as both frequently do. For this is a field where there are, for the moment, no convincing theories or conclusions. There has probably been more discussion of human consciousness in the past 50 years than there was in the previous 10 millenniums. But still we know, to a rough approximation, nothing.

No problem, says McFarland wisely. “Consciousness is a human hang-up.” Or, to put it a better way, consciousness as a scientific concept is a category error. It is a word that more properly dwells in the realm of values and meanings. His question is, “whether aliens (animals and robots) have any kind of self-awareness, and whether the kind of self-awareness that they have could give us good reason to ascribe mental properties”.

This is not as easy as it sounds. McFarland is plainly fond of Border, his irredeemably naughty dog, but, observing its behaviour, he can find no clear evidence of self-awareness. Even the spectacular attainments of chimps and tool- using crows can be explained without dragging in the concepts of subjectivity and rationality that McFarland regards as the essential requirements of self-awareness. This is not to say that alien minds do not exist, merely that the arguments that they do or don’t are as perfectly balanced as Hillary and Obama.

This uncertainty leads to appalling practical problems. For 50 years, scientists have been promising us artificial intelligence (AI). Indeed, some now believe in the imminent arrival of “the singularity”, the moment at which we will build our last machine, an AI computer that will constantly increase its IQ and solve all our outstanding problems. But the prosaic truth is that AI is utterly stalled. One reason is that scientists have been trying to attain some kind of generalised, abstract intelligence. But, as McFarland points out, intelligence that is not about something is not intelligence at all. Evolution provided us with our own aboutness (intentionality is the technical term); we, therefore, have to provide machines with aboutness before we can start talking about intelligence. You can’t aim for intelligence, you can only aim for what you want your machine to do.

In this context, McFarland makes the valuable point that AI is as likely to be developed in the marketplace as in the lab. Market pressures such as price and functionality are more likely to cajole machines into self-awareness than theory. If somebody can make an effective household robot, then somebody else will make a better one and so on. Indeed, one can glimpse — this is me, not McFarland — movements towards at least the possibility of intelligence in the behaviour of cash and washing machines as much as in laptops. And there is a theory that the internet as a whole is already intelligent, a theory that is not entirely absurd.

The big point is the sheer oddity of human self-awareness. It seems unique. Yet our main mistake is to assume that our self-awareness is the last word on the subject. Alien self-awareness may be something utterly different. If a lion could speak, said Wittgenstein, we wouldn’t understand him. This may or may not be right about lions, which, like us, are creatures evolved over billions of years from the stuff of the earth. But it’s likely to be right about machines that share none of this inheritance. The first sign that your laptop is intelligent may be that it appears to turn itself off. In fact, it’s just started thinking in ways you cannot begin to imagine. There is a computer in a science-fiction story by Stanislaw Lem that does precisely this.

None of which may matter in the slightest if we just want machines to do things so well that they seem smart. We can get there without worrying too much about whether they have minds. Lawyers may intervene by saying our smart lawnmower has rights. But, on the whole, if we ditched all our theories of consciousness and closed down the AI industry, technological progress would probably continue unabated.

This is, you will gather, a rich and enthralling book. It is damaged by bad editing. A book on consciousness that has “propriocention” for “proprioception” is asking for trouble. And there are significant passages that either don’t quite make sense or aren’t as clear as they might be. The upside is that it doesn’t condescend or patronise. It gives you the contemporary problem of mind in all its raw complexity. It reaches no conclusions because there are none. There is just the hope that the smart robot when it arrives will be accountable — although whether to itself or us remains, like everything else, unclear.

Explosive consequences

A huge stumbling block for scientists trying to create artificial intelligence is the “frame problem” — how to get machines to work out the consequences of their actions.

To show how tricky it can be, Mcfarland cites a robot called R1, that was told by its designers to get a spare battery from a locked room that also contained a time bomb. Poor R1 located the key to the door easily enough, it even worked out how to recover the battery, by pulling the wagon on which it sat out of the door. What R1 didn’t understand, however, even though it had been told that the wagon also contained the time-bomb, was the consequence of pulling that wagon out of the room. Cue big explosion, and no more R1.

GUILTY ROBOTS, HAPPY DOGS: The Question of Alien Minds by David McFarland
OUP £16.99 pp252

Available at the Sunday Times Books First price of £15.29 (inc p&p) on 0870 165 8585

Times Recommends

• Superfreakonomics by Steven D Levitt and Stephen J Dubner
• The Times picks the 50 best paperbacks of 2009
• Vincent van Gogh’s letters show the soul of the artist