To help clarify what intelligence and morality mean, and whether animals qualify as possessing them, I’m going to take an unusual, but I hope illuminating, detour through some of the setbacks and advances in building intelligent machines.
In his award-winning book On Intelligence , the computing pioneer Jeff Hawkins explains his quest to build ‘real’ as opposed to ‘artificial’ intelligence in computers by mechanically replicating the structure of the mammalian neocortex. Hawkins gives an excellent 20 minute summary of his book in the TED talk below:
According to Hawkins, the quest for intelligent computers got off on the wrong foot from the start when it set out to satisfy the Turing test. Kicking off the quest for intelligent machines in 1950, the British mathematician Alan Turing defined an intelligent computer as one that can fool a human into thinking it is dealing with another human. For Hawkins, Turing’s concept has misled three generations into thinking that intelligent machines are machines which act like people, and that human brains are like computers.
This conception has led artificial intelligence to some hollow victories and some dismal failures. The hollow victories have involved computers beating humans at highly specialised and abstract computational activities, where they are effectively just very high-powered calculators: Deep Blue defeating chess grandmaster Gary Kasparov is the pinnacle example. The dismal failures have been in getting machines to do everyday things that ordinary humans find elementary, like recognizing speech or faces, or in catching a tennis ball.
To explain what makes a Golden Retriever able to do all this where our most advanced computers fail, Hawkins proposes a new theory for intelligence based on how he understands the neocortex to function in humans and other mammals. For Hawkins intelligence is the power to make successful predictions about the world based on memories. The role of the neocortex is to store sensorimotor information in memory, and to continually cross-reference these memories against new experiences to create expectations.
By simulating the hierarchical feedback structures in the mammalian brain, Hawkins believes we can build intelligent machines which make creative predictions and which learn. These mechanical brains will be much more powerful than the human neocortex, both because they can be much larger (as they are not be limited by the size of the human skull) but also because the basic operation of a computer, being purely electrical, is 5 million times faster than electro-chemical transmission of signals through synapses. Hawkins is already making breakthroughs applying this work through his company Numenta.
But what is particularly interesting to us is what these intelligent machines won’t be able to do: they will not have the emotions, desires or motives of humans, because as Hawkins explains, “the human mind is created not only by the neocortex but also by the emotional systems of the old brain and by the complexity of the human body. To be human you need all of your biological machinery, not just a cortex.”
For this reason, neither moral nor immoral robots will be arriving on the scene anytime soon, which should reassure anyone fearing a Terminator or Matrix style machine-uprising! On the contrary, lacking the emotional structures for shame or boredom will enable machines to overtake many of the tasks we find unbearably humiliating or insufferably tedious, as well as undertaking complex tasks in alien sensory environments and abstract mathematical realms.
On the down side, this means digital immortality is an equally distant prospect, as we won’t be able to ‘upload’ our personalities to a computer until we can completely reverse engineer the human body, a technological feat which will not be realised anytime soon.
Unlike machines, however, many animals share “the emotional systems of the old brain” with humans, and an important subset of these have also developed neorcortical intelligence. Until very recently, it was believed only mammals had evolved this brain structure, but latest research has found that a huge proportion of the brains of songbirds is taken up by a cortex like structure.
The hypothesis we explore in Moral Menagerie is that the most intelligent social animals will have something approaching morality and sentience. Far from being a novel hypothesis, it is one we share with the father of evolutionary theory, Charles Darwin himself, who, in The Descent of Man wrote:
“The following proposition seems to me highly probable – namely, that any animal whatever, endowed with well-marked social instincts, would inevitably acquire a moral sense or conscience, as soon as its intellectual powers become … nearly as well developed, as in man”