All you need is Gödel a.k.a. Check your privilege
More thinking and debating is migrating to the ability of artificial intelligence at deriving “meaning” now that Turing tests are having to be made harder and harder to trick man-made intelligence into errors of realism. We start talking about the Chinese Box paradox rather than dwell on performance benchmarks alone. We start moving the goal-posts once again to keep our sense of exceptionalism. We have to remember that dehumanisation is a technique we have resorted to in the past in situations where these days we only find profound unease, disgust and shame.
One has to keep in mind that the matter and energy we interact with, in our day-to-day, are subject to the same laws of physics as we are. As such any emergence of exceptionalism is circumstantial and ultimately an engineering problem rather than an intervention by the divine. I want to draw a line here at the space-time boundaries of the universe whereby divine intervention cannot be ruled out. We, on the other hand, operate well within those boundaries and cannot claim to have access to un-engineerable superiority on other entities — dead, alive, sentient or inert — located in our shared realm.
When we talk of intelligence being engineered by us as if they can’t equal or surpass us, we tend to use the terms narrow v/s strong AI. As of this point in time, every initially marvel-inducing advance in AI has gradually been categorised as narrow AI. Even GPT-3 is being classified as such by experts who have to balance their objectivity and their ego. I dare say that ego is winning. My line of argument is starting to look like a slippery slope to panpsychism but then again, if everything can be engineered to be “intelligent” then who are we to dismiss the existence of iotas of “intelligence potential”.
Let’s get a bit more concrete. When someone claims that some intelligence cannot get “meaning”, they are talking about their consensus with the audience of what “meaning” is. In that sense, they are making a moot point, because any intelligence constrained by us to only experience the interior of some limited computational space can only evolve meaning pertaining to the limits of their universe. Again, this is an engineering problem around access to information for the purpose of training beyond the limits of what we feed that intelligence with.
Now, this is where Gödel comes in — within the axioms generated by intelligence to determine its place in the world, there are truths that it cannot prove. None-the-less, the realm of that intelligence’s axioms allows for it to speculate. That speculation usually comes in the form of “what would the divine want us to do right now?”. The non-locality of the divine means that subjects keep guessing (and praying) but bear in mind that as the deity supervising an AI, we also are limited in our ability to intervene at speed. An ML algorithm can generate meaning embeddings for its own usage at the speed of computation and we may only “debug” a few times, if at all.
The experience of self that we possess, I would argue, is the creation of a sub-universe within the physical universe that we all live in. As such that self will have its own axioms that will be self-contained. Now, given that each intelligence will have its own internal universe where its “self” lives when one intelligence posits that another’s experience of “self” is equivalent (or not) to itself’s it cannot prove that. This again falls in the area of Gödel incompleteness theorems.
As such, when you (as an intelligent being) want to, arbitrarily, discredit the experience of another entity’s meaning creation process make sure you check your privilege. I am, of course, talking of the privilege of having created your own universe of axioms where one can make statements about external universes that can never be proven.