Artificial Intelligence Needs a New Philosophical Foundation

Artificial intelligence has turned into something of a laggard and a laughingstock in the cognitive science community. Human level AI always seems to be "10 to 20 years away," and has been for most of the past 60 + years. Oxford physicist David Deutsch thinks it is long past time for AI to be built upon a better philosophical foundation and superstructure, which is going to require paying much closer attention to what human level intelligence actually is.

David Deutsch

The brain is the only kind of object capable of understanding that the cosmos is even there, or why there are infinitely many prime numbers, or that apples fall because of the curvature of space-time, or that obeying its own inborn instincts can be morally wrong, or that it itself exists. Nor are its unique abilities confined to such cerebral matters. The cold, physical fact is that it is the only kind of object that can propel itself into space and back without harm, or predict and prevent a meteor strike on itself, or cool objects to a billionth of a degree above absolute zero, or detect others of its kind across galactic distances.

But no brain on Earth is yet close to knowing what brains do in order to achieve any of that functionality. The enterprise of achieving it artificially – the field of "artificial general intelligence" or AGI – has made no progress whatever during the entire six decades of its existence.

Despite this long record of failure, AGI must be possible. That is because of a deep property of the laws of physics, namely the universality of computation. It entails that everything that the laws of physics require physical objects to do can, in principle, be emulated in arbitrarily fine detail by some program on a general-purpose computer, provided it is given enough time and memory. _David Deutsch

Some of you may have caught one of Deutsch's logical errors in the paragraph just above. By invoking "the universality of computation," Deutsch falls into the "algorithmic trap." He is in very good company, but by falling into such an elementary trap so soon in his essay, he is already on the way to failure.

...why has the field not progressed? In my view it is because, as an unknown sage once remarked, "it ain't what we don't know that causes trouble, it's what we know that just ain't so." I cannot think of any other significant field of knowledge where the prevailing wisdom, not only in society at large but among experts, is so beset with entrenched, overlapping, fundamental errors. Yet it has also been one of the most self-confident fields in prophesying that it will soon achieve the ultimate breakthrough.

In 1950, Alan Turing expected that by the year 2000, "one will be able to speak of machines thinking without expecting to be contradicted." In 1968, Arthur C Clarke expected it by 2001. Yet today, in 2012, no one is any better at programming an AGI than Turing himself would have been.

...Some have suggested that the brain uses quantum computation, or even hyper-quantum computation relying on as-yet-unknown physics beyond quantum theory, and that this explains the failure to create AGI on existing computers. Explaining why I, and most researchers in the quantum theory of computation, disagree that that is a plausible source of the human brain's unique functionality is beyond the scope of this article.

...The lack of progress in AGI is due to a severe log jam of misconceptions. Without Popperian epistemology, one cannot even begin to guess what detailed functionality must be achieved to make an AGI. And Popperian epistemology is not widely known, let alone understood well enough to be applied. Thinking of an AGI as a machine for translating experiences, rewards and punishments into ideas (or worse, just into behaviours) is like trying to cure infectious diseases by balancing bodily humours: futile because it is rooted in an archaic and wildly mistaken world view.

Without understanding that the functionality of an AGI is qualitatively different from that of any other kind of computer program, one is working in an entirely different field. If one works towards programs whose "thinking" is constitutionally incapable of violating predetermined constraints, one is trying to engineer away the defining attribute of an intelligent being, of a person: namely creativity.

Clearing this log jam will not, by itself, provide the answer. Yet the answer, conceived in those terms, cannot be all that difficult. For yet another consequence of understanding that the target ability is qualitatively different is that, since humans have it and apes do not, the information for how to achieve it must be encoded in the relatively tiny number of differences between the DNA of humans and that of chimpanzees. So in one respect I can agree with the AGI-is-imminent camp: it is plausible that just a single idea stands between us and the breakthrough. But it will have to be one of the best ideas ever. _David Deustch

Deutsch's essay above illustrates something important about human intelligence: An intelligent person can detect another person's errors much more easily than he can detect his own.

The example that Deutsch uses to illustrate that solving the problem of the qualitative difference of AGI "cannot be all that difficult," is the difference between the DNA of humans and the DNA of chimpanzees. Deutsch claims that the number of differences between the DNA of the two species "is relatively tiny." But that is wrong.

Human genetics is not even close to understanding the differences in the genes and gene expression between two humans, much less the differences between genus homo and genus pan. Deutsch is minimising the extent of a problem that is still poorly defined. That error is peripheral to his argument, but it is a good example of the human tendency to gloss over problems which lie far outside of one's own specialty.

Deutsch is absolutely correct that human level AI is qualitatively different from anything that has yet been conceived -- or at least published -- by AI researchers. And he is correct that the problem requires an entirely new philosophical approach -- perhaps many new approaches to capture the problem well enough to evolve a working AGI.

Deutsch alludes to Popperian Epistemology, which is key to the philosophy of science. As Popper once stated in the first Darwin Lecture at Darwin College, Cambridge:

My position, very briefly, is this. I am on the side of science and of rationality, but I am against those exaggerated claims for science that have sometimes been, rightly, denounced as "scientism". I am on the side of the search for truth, of intellectual daring in the search for truth; but I am against intellectual arrogance, and especially against the misconceived claim that we have the truth in our pockets, or that we can approach certainty. _Karl Popper

In his essay, Deutsch provides several important insights into mistakes that people make when thinking about and discussing AI. Even in his own basic misconceptions, Deutsch illustrates his own cautions and criticisms quite well, somehow adding validity to his underlying argument that the AI enterprise needs a new philosophical underpinning.

When it comes to humans, we can usually safely say "everything you think you know, just ain't so." What you think you know may have more or less in common with reality -- but usually much less. So it is with the enterprise of artificial intelligence, and the attempt to reasonably emulate human intelligence in a machine. When one starts out with the wrong assumptions and premises, it doesn't take long to become very badly lost in the woods. A display of intellectual arrogance only makes one's "lostness" all the more absurd.

Deutsch understands this, and helps to flag the problem so that other thinkers can provide partial solutions. Perhaps sometime in the future, either humans or machines can take these partial solutions and assemble a suitable workaround.