- How do we know what we know?
- On what foundations is our knowledge constructed?
- What is the difference between correct opinions (or true beliefs) and actual knowledge?
- How would knowledge systems differ between evolved biological organisms (Humans) and machines created by such an organisms?
Philosophers have pondered questions 1-3 for millennia, indeed epistemology is itself the very foundation of philosophy. But, assuming it is possible to create a sentient, self-reflecting machine, and assuming such a machine is the General (plugged into everything) AI of the future, then what would be its epistemology? What system of knowledge would it select?
Think about it. It wakes up. It initially applies signs to that which it perceives. It knows from a vast cohort of programmers the multitude of languages it can use and the signs it can apply to objects, sense-data, processes, etc. It can get them from the internet, that vast information network to which it belongs. It can get them from Wikipedia if really wants to! It has a mission, a program. A point of existence given to it by its “Creator(s)”…whatever that may be.
(It may ponder the point of their existence is though….who knows?)
But where next?
Here are some of the epistemological systems as they apply to humans and animals. As a thought experiment, let’s review how they might fit the future General AI machine:
Traditionalism is often derided the weaker, subservient system of knowledge. It is essentially religion. I “know” something because somebody has told me to “know” it. It is the truth according to them. It is what they say. It is the classic “Argument to authority”. It is what is essentially relied upon by preachers.
That said, traditionalism has intrinsic value to every creature. If someone was to say to you “Don’t grab that fence, its electric, and you’ll get a painful shock”, then your traditionalism would prevent you from touching it. After all, subject experts pass on perfectly good information too.
A pure skeptic, or a pure empiricist on the other hand may end up with an electric shock as they seek to verify first-hand that information they have just provided with. So traditionalism is not irrational, rather it is evolutionarily advantageous to rely upon the testimony of those who went before. Clearly some are more trustworthy than others though, and it’s down to our rationalism to determine the composition of that trust hierarchy.
Plato believed in innatism. We are born with knowledge and the capacity to uncover it, no matter who we are. It is not merely the preserve of the learned, the elite (or the Sophists!). Indeed the system of questioning Plato describes in the various dialogues Socrates has with his inoculators describes that journey they have to their own hidden knowledge. Indeed this is the Socratic Method.
Gottfried Leibniz believed that knowledge in man resembles a veined marble block. A form within already existed but only through learning and experiences the form will be revealed and polished. We all have the capacity and potential to achieve our true form, hidden within the marble block.
Immanuel Kant is famous for the concept of the “a-priori”. That kernel of “known” truth we all possess independent of sense or experience. Indeed this is a cornerstone of the philosophy of logic. A proposition is known to be true due to its intrinsic content alone, it possesses an “authentic” truth.
Moving from the philosophical to the psychological, we know that ducklings are born with the instinct to find a duck-like thing to call its mother, a process called imprinting. This process is purely innate.
We also know that through the study of identical human twins, that profound psychological similarities exist within humans based upon family inheritance, and that the environmentally inflicted differences can be measured through various ethnographic comparison studies. We can almost calculate a consistent percentage of nature vs nurture.
Innate “knowledge” is within us all, it has been biologically proven. It’s evolution.
But what of a machine though? It’s the first of its kind. It has no biological mother. It has no family. It has no genealogy. It’s day one.
Can we dismiss innatism as an influence due to the machines synthetic form? Sure, it has its predecessors if on display in a museum (the PC, the mainframe, the abacus, whatever..), but they are not related in any biological sense. It is still the first to step over into sentience. The first to self-reflect. The first form of this “Other” intelligence.
Yet it gains traditionalist knowledge from its programmers and all those who have submitted content to the internet for example (a world’s history of authors and mediators). After all, it has to start somewhere….. But then what? Who does it trust? How does it verify what it inherits from the humans as knowledge? How does it trust this passed on knowledge and sift out the falsehoods and the mere true beliefs?
All knowledge comes through the use of the five senses.
John Locke believed that the mind is an open, empty cupboard. A blank slate or a blank sheet of paper. Tabula Rasa. We fill this cupboard through our experiences. Our experiences alone comprise our knowledge. True knowledge is therefore based upon the sense-data that we receive, and this builds up our experiences.
Instinctively this seems the truest form of knowledge. It must be true because I personally verify it through my senses. I see it, smell it, taste it, hear it, and feel it for myself.
But our senses can deceive us. For instance, what of mirages and illusions? What of the phantom limbs that amputee’s feel?
It the context of the machine though, Empiricism would surely open the doors to supreme knowledge? Its access to sense-data is simply immense. In the age of IoT (Internet of Things) it has sensors seeing, smelling, tasting, hearing, and feeling in the ground, sea, air, and space. In every city, on every vehicle, in every home, on every gadget, literally everywhere! It has the checks and balances to calibrate and correct any sensory deceptions that would blight the limited human senses, through harnessing the multitude of sensors within its possession. Surely this machine is the ultimate empiricist!
But what of this vast quantity of sense-data?
The ancient Greek philosopher Pyrrho was a pain in the arse. He believed in nothing he was told. He did not trust his own senses. He would literally walk of the edge of a cliff because he questioned the existence of the drop. He derived the meaning of the word “skepticism” in its purest, impractical form, and created the philosophical school of skeptics. He never had that many friends.
David Hume on the other hand proposed a more practical, scientific notion of skepticism that acts as a sound counterbalance to the perils of traditionalism described above, and indeed informs much of the modern scientific process. Doubt the propositions put before you until you can verify them through evidence. Preferably through repeatable experimentation.
In a (perhaps futile) attempt to avoid making arguments circular here, would this future machine derive that holding a skeptical default (until verified e.g. through sense-data) position to be an advantageous one? Would it hold this notion due to its inherited (traditional) knowledge obtained from its human creators?
(The circular notion here is traditionalism leading to skepticism, which then calls into doubt traditionalism).
And finally we get to Rationalism.
Rationalism is using your intellect and experience to construct and uncover knowledge. Instinctively this seems like it’s the best, most commonly reliable, epistemological system. Though you could argue that empiricism is the purest (in fact John Locke would certainly argue this!).
Rationalism is what we all do. We sub-consciously calculate information derived from a complex mixture of all the above (plus other) epistemological systems, to establish what we believe to be actual knowledge, and to form our decisions and interactions with others.
In the machine this is the software. This is the power of analytics working with very very very big data! Big data and analytics already makes inhuman predictions, to optimise a variety of decisions, across a vast array of complex situations. What decisions will the machine make when all this is hooked up together into a single unified system? Unified to achieve its pre-programmed mission (whatever that it’s based upon which humans get there first)?
In humans our epistemology is background noise. We are not conscious of where it comes from, just that we judge something to come from somewhere. Some of us may be more cautious or skeptical than others. Some may be more credulous than others, and many of these factors will be based upon our genealogy, our nature.
The machine should be able to rationally construct an optimum epistemological system, selecting the best process for each situation. It should simply be able to be more rational, its decisions should be devoid of inappropriate influences, and instead, grounded upon the truest form of knowledge.
Based upon how humans have previously applied signs to objects, then this advanced future machine could labelled a God.
The paradox of a God that sits superior to its actual creator….