Euclid Kills The Ego, Doing Using and Interacting, and Face Recognition
There is no "I” in "postulate”
The opening of Michael Polanyi’s book The Tacit Dimension outs the whole work as a glorified instance of motivated reasoning:
I first met questions of philosophy when I came up against the Soviet ideology […]
Not that there are no good intuitions in there, far from it. To the contrary, it is certainly a book worth reading.
Still, Michael has an axe to grind.
And it seems to me that he is fighting against everything that I have always held -since my college days- to be true and holy1. And that is that the mission of the scientist is precisely to remove those elements of the unspoken that make theoretical work dependent on implicit assumptions2.
What makes Euclid great is not so much the theorems, as the idea that he was willing -and able- to condense what we are expected to believe into a few simple sentences: postulates. Theory is about boiling down inevitable belief to a minimum, with the added bonus of making it impersonal. I do not have to apprentice with a geometer to learn geometry, because my learning is not about him, it’s about It.
The sentence
Our message has left something behind that we could not tell, and its reception must rely on it that the person addressed will discover that which we have not been able to communicate
is a long way to spell failure, at least after 300 BC.
The economic literature spells it DUI, instead. It stands for Doing, Using, and Interacting and it is usually regarded as an alternative (complementary?) mode of innovation to STI, which stands for Science, Technology and Innovation. Implicit versus codified knowledge.
Obviously the overlap between these concepts is not perfect, but it is certainly interesting that the inability –or unwillingness- to take oneself out of the equation has strong psychological and even political implications. For instance, the only social compact that is standing up to the looming tide of technocracy -which is a way of misunderstanding the content of the scientific enterprise rather than rejecting it- is predominantly made up of small entrepreneurs, professionals, artisans. People who are making a banner of the illegibility of their own know-how, which makes them irreplaceable.
Irreplaceable for now, at least, which brings us to AI. Polanyi’s chief example of tacit knowledge is… drum roll… face recognition:
We know a person’s face, and can recognise it among a thousand, indeed among a million. Yet we usually cannot tell how we recognise a face we know. So most of this knowledge cannot be put into words.
Until recently, tacit meant impossible to store outside of a human being’s skull. Coding knowledge into symbols –paintings, words, Jupyter notebooks, etc.- was the only way we had to store it outside ourselves. To be able to survive and reproduce outside of their biological host, ideas had to pay this price: to be turned into symbols that humans could understand in isolation. An unavoidable case of lossy compression.
That is no longer the case. Since at least the Viola-Jones algorithm from the early 2000s, the tacit knowledge of face recognition has been able to exist outside of a biological substrate in a form that does not make it any less tacit. It can be easily stored and transmitted without the need to translate it first into some kind of low-bandwidth representation. This was precisely the stumbling block that lead to GOFAI approaches crashing and burning, resulting in the AI winter that ended a few years ago.
The stumbling block is gone, now. Storage and transmission of knowledge are possible without the unbridled lucidity that the author of The Tacit Dimension feared would destroy our understanding of complex matters3. So we can finally have the worst of both worlds: knowledge that is still personal, just owned by someone else.
This to the point that upon my first reading of The Tacit Dimension I scribbled in the margin If someone proved to me that Polanyi is with the truth and that in reality the truth were with Polanyi, then I should prefer to remain against Polanyi rather than with the truth. Tongue in cheek to be sure, but still.
Clearly, meta-theoretical work, that is the job of improving on existing theory or ditching it altogether for a new one, is harder to free of implicit assumptions. Squeezing out the toothpaste becomes harder the nearer you get to the end of the tube.
It is interesting how the students who could easily swallow the lack of clarity and rigour that characterised most early lectures in introductory physics courses were also the ones who were most quickly convinced that they understood everything…