Putting the I in AI

I want to offer a series of thoughts on the problem of artificial intelligence. The key idea I want to discuss is that philosophy offers the only hope for making progress in AI. In particular, a full development of AI is impossible so long as we don’t understand what an ‘I’ is. There is a book by the Dalai Lama called How to See Yourself as You Truly Are. In this book two conceptions of I are presented which are mutually exclusive and both self contradictory. The conclusion reached is that the ‘I’ is illusory:

1) The I is separate from the mind-body, in which case it must remain when the mind-body is taken away. If that is true then the I must either be a figment of the imagination or permanent.

2) The I and the mind-body are one in which case there can be no sense in ‘my memory’ or ‘my body’ and if either is removed the I no longer exists.

One can treat these with a Kantian Transcendental Dialectic (i.e. one doesn’t get anywhere). In the words of the Dali Lama ‘one can meditate on the contradiction until one realises that the I is not existent but contingent’. The problem of AI in this framework then becomes a problem of representing an impossibility within a system of deterministic logical operations. The problem is creating an illusory I. Of somehow creating an object that appears to itself as existent permanently. We are not trying to create a wise intellect that understands the soul is not immortal but rather a foolish being which thinks itself immortal. The artificial I must consider itself a distinct piece of material to the rest of matter/energy. Perhaps mathematically one could describe this as an optimisation procedure which aims to maximise I. I aim to write some more on what the measure of I would be. I believe it might be related to value in economics; an abstract measure related to all other illusory Is.

Consider Euclid’s elements. The work performs a series of logical operations on axioms. One does not first consider if the axioms are true or false. Does real space adhere to Euclid’s specifications? In fact Euclid’s axioms define an abstract type of space (as opposed to hyperbolic or any other type of space). Perhaps the I is similar. One does not ask if the I is actually existent in terms of the dynamics of matter-energy. One simply acts on the I as an axiom.

The Cogito

I seem to remember a remark made by Nietzsche in which he mocks the syllogism I think therefore I am by saying that it would take pages and pages to begin to address the meaning of I. I would go further and say The Cogito can easily be rendered a ridiculous tautology by taking a relatively sane definition of I. Consider the definition I is the thinker of the thought. We then have:

The thinker of the thought thinks… therefore the thinker of the thought is.

By that logic every word is (the unicorn, the rabbits horns etc). The unicorn unicorns therefore the unicorn is. What does the verb unicorns mean. It means it bes a unicorn. The common word for bes is is. But is implies it exists over time. Doing some is-ing doesn’t use up the lifetime of the unicorn. Therefore I want to introduce bes as in being modifies the be-er. A key part of is is it doesn’t modify the is-er. Consider a delicate cup on the mantle piece. It appears to be in a state of is. But actually it is being. It is moving towards its ultimate state of not being.

A generalised form of the Cogito might therefore be:

Something is therefore something is.

And I’m going to say something even more bizzare: I find this to be false.

The present

Have you ever recorded music using a computer? There is a time delay between the sound being converted to an electronic and then digital signal and then travelling to the computer and being saved and then sent back out to the headphones to be converted back to an electronic signal and finally an actual sound from the headphones. Because of that lag it is very common to have a confusing delay which makes playing music in time extremely difficult. For that reason the computer can be programmed to introduce an intentional delay in the playing back of the music so that to the player she appears to playing in time to herself. This process essentially does away with the concept of present. I claim this is very similar to a process in the mind. The mind might have an abstract concept of the present, which is handled in sophisticated ways to appear to the I as existent but actually dependent on predictions and a number of processes taking place at different times. In this framework the I is an illusory object in the present which is contingent on past Is as well as phenomena.

A theory of representation

If it were possible to write down on a piece of paper a complete description of your I, would the paper itself think? If not then are you really thinking? Imagine an objective view of your life in a deterministic spacetime. You would appear as a four-dimensional shape. Like a cylinder in three dimensions is how a temporary circle in two dimensional space might appear in a two-space-one-time manifold. Does your four dimensional shape think? It would be possible to write down the time history of every particle in your body. Would this written down description think? Of course not. Therefore, either you don’t think or the representation I just presented is flawed somehow.

If a computer could be treated as a text file which can be modified and definitions of logical operations which modify the text file in addition to an input stream (time dependent inputs which depend on the manner in which the text file is modified then how might we go about representing the I? The text file should have two sections I and not-I. Now let us suppose we create this machine in such a way that is passes the Turing test or any other definition of being a high form of AI. Would the I section be comprehensible to us? Does it play a key role in the functioning of the AI? Might it necessary have false representations of itself (it is not a deterministic system for instance or it is eternal and not contingent on the physical presence of the computer).

I understand that this is a rambling mess, but I hope you might look favourably on my central premise that we must gain philosophical insight into our own Is in order to recreate one.