The most incredible aspect of Deep Learning to me is that it showcases how one can make surprising progress when the premise is that of freely admitting ignorance.
We really don’t know how the human brain works stuff out. Sorry. But, here’s a pretty good guess. Let’s imagine a neural network, and let’s figure out nifty ways to inform it with a bucket-load of data that we have been storing for a while, without really knowing when and how we might make use of that either.
Naturally, the algorithms involved in making the AI smarter, the data repositories more reliable and their communication more seamless is where the real heavy-lifting is happening these days, and the pace of progress is truly staggering.
However, the point is that the premise was one of ignorance, though, of course, a very special kind of ignorance. An ignorance that allows taking unorthodox approaches that are creative and exciting. It is an ignorance that is informed by the idea that technology will eventually catch up and open up possibilities for applications both imagined and novel. Such deliberate ignorance sometimes permits creative insights because it forces divergent thought processes.
What has all this got to do with transactions? A lot, as it turns out. It really bugs me when the word “transaction” is used as though it were easily and broadly understood to be some universal concept. It just isn’t. It means very different things to different people, and sometimes even to the same person depending on the context. Withal, some transactions are antithetical to deliberate ignorance.
Before I develop this idea, I want to state my own understanding of what is a transaction. A transaction is an exchange of a package of rights. The intensity of the component rights immanent in any given object can change, but they are, principally, just two: the right to use and the right to exclude usage. You get a variety of types of goods when you tinker with the combinations of those rights, but that is, in a nutshell, the gist of it. There are subtle nuances in this definition, by the way. For example, one can think of other rights, such as those pertaining to the stream of income generated from usage and to share or trade fractions, but these need not concern us here.
Now, when we parade an example of a transaction such as Alice pawning off a thingamajig to Bob in exchange for some compensation, we are making the implicit assumption that Alice understands the rights over the object she has and has the ability to transfer the whole package associated with the t-jig. And we assume that Bob, too, understands what the rights are he is getting, and will feel well pleased with his new acquisition. All this is usually shoved under the carpet when we say that the t-jig’s price stands in as a unidimensional proxy for the package of rights it comprises.
Here’s the rub, though. It’s no good saying that a million people will stand as witnesses to the fact that Alice did actually sell the t-jig to Bob if, somewhere down the line, Bob sells the t-jig someone else and finds that he is no longer in control of the whole package of rights, just as had before. Everyone might agree that Bob has the t-jig. Indeed, his account book and the account books of each of the million witnesses might all attest that Bob indeed did fork over his hard-earned and got the t-jig for it, but this assumes that the package of rights is itself immutable over time and use.
Very often the package is immutable and “transactions” purr along chirpily. I believe that this defines most transactions, and has, therefore, blindsided us in those cases where it doesn’t hold true: cases that comprise deliberate ignorance.
When Bob uses the t-jig, excludes others from using it, or some combination of both, the transformation to the inherent package of rights can either be affine or it may not be. When the transformation is affine the package is deformed in the transaction, but it remains tractable. .
A t-jig may be added to a u-jig and a v-jig may then be removed from the result, but, in the end, so long as the w-jig that remains is also some object — from a broad class of objects — that can be conveniently defined by a package of rights that the t-jig held, then, for all intents and purposes, the transaction is pretty straightforward. The price proxy works perfectly well here; our brain is happy to work linearly, but does remarkably poorly when taxed with non-linearity in situations
Troubles arise, however, when the transformation of the package of rights is non-affine. This is how I think of what the heck an externality really is; the rights over the component goods are morphed and moulded in such a manner that the resulting thing requires an addition or subtraction of a hypothetical or accidental good (that represents the balance of the rights) in order to make it commensurate with the goods involved in the initial exchange.
Deliberate ignorance has to do with an inability or unwillingness to pin down the nature of the package of rights that inhere within this hypothetical good.
Regulation deals with deliberate ignorance (mostly, it is just that) almost exclusively; the inability, that is, to define the hypothetical good precisely without mucking around with the underlying good itself!
And, the more squarely innovation concerns itself with new and novel applications for unorthodox combinations of ideas, the more likely it is that deliberate ignorance will play a role. Since the line between those transactions that are affine and those that are not is much harder to establish when ideas are concerned, externalities are especially rife.
A limited view of transactions distorts what innovations can arise, since they must then be congenial to our view of what a transaction ought to be.
But, in an age of ideas, this is not a concern that spells inexorable defeat for the intrepid entrepreneurial mind. A key benefit of platforms for the sharing economy, for example, is that they take on the charge of representing an affine package of rights, while, simultaneously, permitting their separation and enabling non-affine transactions of objects within. The platform’s value, in a sense, is that it becomes the hypothetical good I spoke of earlier. This is also a key benefit of linking a blockchain architecture with a cloud service for applications that are expressly about permitting access to ideas for generating innovations.
The point is that deliberate ignorance needs to be jealously protected whenever the potential for innovation looms large, and we ought to have a more nuanced view for what a transaction can be.