17 October 2022

We know nothing.
That’s the only thing I’m sure about.

I was reading Feynman’s Lecture on Physics to gain some insights on the fundamentals of quantum physics and two things came clear to me.

First, Richard Feynman has climbed to the top of my list of awesome human beings to have ever lived on this planet. (That position changes with alarming frequency, I must also confess.)

And second... Wave-particle duality?!?! Really?!?! C’mon!!!! 😡

I was spoonfed with wave-particle duality in my university years and I took it at face value, giving it no second thought whatsoever. One more concept to get my degree. But… really? That thing just doesn’t make any sense to me today.

And Heisenberg’s Uncertainty Principle? Really!?! 🤬 I call that cheating!!! When I was a kid and I played with my little sister, I would come up with random rules to make the game interesting or steer it to my advantage. Like:

Nope! I’m sorry, that goal doesn’t count, the ball needs to touch the floor first.

That’s Heisenberg:

Nope! I’m sorry, you cannot really know the position of a particle with that level of precision.

via GIPHY

It got me thinking on Relativity too. Gravity is the curvature of space time?!? What the f#ck is that too?!?!

Look, at the risk of showing my limitations here, this is what I came up with:

On a scale from the biggest to the smallest, I’m able to understand just a fraction of it—below the solar system and above the atom, to be specific. I would call that range the human scale, a representation of my observable reality (I know I’ve never seen a hidrogen atom but at least I can picture it in my head).

Beyond those limits it’s just numbers and models. I’m stuck with Newton—or extended Newton let's say. Einstein and Feynman went over my head. Yes, I know you can math the shit out of it, but I call that the infrared/ultraviolet of my intellect… I can deduct they exist but that light doesn’t shine on me, I just don’t grasp it. And you don’t grasp it either, don’t lie. 🫵

The Human Threshold

The human scale defines other relevant limits. I explored them recently when talking about IT Production Management— ✋ WAIT! Don’t close this window yet! What I’m about to say applies beyond production management into more general process management, if it’s any of your interest!

Check this time scale, equivalent to the size scale I used before—I've deliberately left the higher range (years, centuries, etc) out of scope as you don't really want this post to be longer, trust me.

Let’s use the time scale to map the cycle time of some processes we typically run at banks, like monthly accounting closing, daily publication of VaR and P&L or almost real-time trading.

The time scale also has human limits to it, in this case defined as the minimum cycle time in which human intervention is possible. If a transaction must be processed in the space of miliseconds, one cannot expect a person to perform any step in the middle

The human threshold is here, somewhere in the minutes/hours range, depending on the type of process—look, this is not going to be Feynman level of precision, ok? I might fall short on more things than the Nobel and the hair.

Processes that happen above the human threshold benefit from adaptability, flexibility and intelligence, which are features of most of the human beings I know—most 😉.

Processes below the human threshold benefit from efficiency, scalability and predictability, which are traits of machines—unless when they are programmed by… 🤐

In order to achieve the benefits on the right, we rely on structured data and programmable logic, i.e. computer science.

In order to achieve the benefits on the left, we rely on intelligence, creativity and problem solving.

The things on the right are brand new to the human kind. So brand new that I, at 45 years old, feel myself a witness of every single step that humans have taken below the threshold.

The things on the left we’ve done forever, but I find it fascinating how we continue developing our understanding of human cognition and collective intelligence, and the possibilities that could bring.

Anyway, now that we’ve defined—broadly 🙄—the human threshold, I can state the first principle:

You must accept this by definition, actually—yes, human intervention will be required by exception when things don’t go as expected, that’s how payment remediation queues are brought into existence.

Above the threshold

But it is not the behaviour below the threshold that I find most interesting, it is the behaviour above. So here’s my second principle:

Yes. Let that sink in for a minute.

I said INEVITABLE.

Human beings will intervene in a process if the cycle time is above the human threshold. They will do it even when it is not necessary. People will come up with an excuse to do it, just because they can, just because things happen in their timeframe. Humans will conspire to make their mark. Someone will propose adjustments to the accounts. Someone will perform some manual controls on the VaR data. Everything will be designed to be supervised by a human being, and the human being will intervene, eventually. In doing so, humans will bring in not just intelligence, adaptability and flexibility but also inefficiency, unscalability and unpredictability 😬.

True STP (straight-through processing) can only happen below the human threshold.

Accepting this premise leads us to the following:

All deterministic processes must be designed with a cycle-time below the human threshold.

Accounting is a deterministic process. It’s wrong and anachronic to do monthly closings. We should be able to produce financial statements in real time. There should be no daily batch processes, either. Yes, I know there are many boundary conditions that force us to do it this way, but those boundary condition are also anachronic and that’s the challenge of true digitalization.

True Digitalization

Anything else is marginal optimization. Applying RPA to optimize a monthly process is marginal optimization. It’s an attempt to bring the benefits of the right side into processes that run on the left side. Not good enough.

Using AI to extract the relevant data from a syndicate loan agreement is marginal optimization, too. The true value of AI is to bring the benefits of the left side into the right side, but, for that, our processes must run on the right side to begin with. Once there, we can feel the pain and limitation of relying only on structured data and programmable logic. AI’s challenge is to break those limits and bring some human features below the human threshold, and that’s awesome and scary in equal terms

equal terms?

🤔

that’s another post

and this one’s long enough already.