Human pilots and autopilots need to be trusting friends.

Sir, last year, flying to Washington DC, in a Boeing 337 MAX 9, the clouds were so low that the pilot missed the first approach to land at Dulles International Airport. We passengers were instructed: “Please, please completely turn off all, absolutely all your computers, iPad cellphones and similar, even if closed or placed in air-mode. We cannot afford any type of interference. The pilot has decided he can’t land the plane, and so he will have the auto-pilot do it.” 

Very long minutes of praying and checking up on co-passengers’ cellphones ensued. Finally, without having been able to see any land, thanks to God we landed.`` I asked my wife: “If they have to use the auto-pilot when things get really hard, why do they have pilots at all?”

Tim Harford answers that question in “Of top-notch algorithms and zoned-out humans” FT February 1. Harford writes: “A storm that blocked the sys­tem’s air­speed instru­ments with assist­ive fly-by-wire sys­tem that trans­lates the pilot’s jerky move­ments into smooth instructions. This makes it hard to crash an A330. But para­dox­ic­ally, it means that when a chal­lenge does occur, the pilots will have very little exper­i­ence to draw on as they try to meet that chal­lenge.”

What does it indicate as a solution? Clearly humans and autopilots need to communicate faster and clearer about who is more qualified to be in command. And that means they need to be able to trust each other much more.

Sir, after the experience I described, even though I have often complained about pain over my eyes when landing with a Boeing 337 MAX 9, probably because of some issues with its air pressure control, sincerely, after that landing, tell me how could I ever complain about that Boeing plane?

@PerKurowski

Like algorithms humans can also produce peculiar and unjust decisions, and be almost just as faceless.

Sir, Gillian Tett writes: “as institutions increasingly rely on predictive algorithms to make decisions, peculiar — and often unjust — outcomes are being produced.” “The tragic failings of faceless algorithms”

Indeed, but humans are also capable of producing peculiar and unjust decisions.

What could be more peculiar than regulators wanting banks to hold more capital against what by being perceived as risky has been made innocous to the bank system, than against what, because it is perceived as safe, is so much more dangerous?

And what is more unjust than because of these regulation allowing easier financing to those who want to buy houses, than to those entrepreneurs who are looking for a possibly life changing opportunity of a credit. 

Ms Tett quotes mathematician Cathy O’Neil’s Weapons of Math Destruction with: “Ill-conceived mathematical models now micromanage the economy, from advertising to prisons,” she writes. “They’re opaque, unquestioned and unaccountable and they ‘sort’, target or optimise millions of people . . . exacerbating inequality.”

Well “opaque, unquestioned and unaccountable” that applies equally to the bank regulators who do all seem to follow late Robert McNamara’s advice of “Never answer the question that is asked of you. Answer the question that you wish had been asked of you”

And on “exacerbating inequality”, the regulators de facto decreed inequality

https://subprimeregulations.blogspot.com/2016/03/decreed-inequality.html

PS. An aide-mémoire on the major mistakes with the risk based capital requirements

@PerKurowski

We would all benefit from algorithms tempering our bank regulators’ human judgments.

Sir, Sarah O’Connor, discussing the use of algorithms when for instance evaluating personnel writes: “The call centre worker told me the software gives lower scores to workers with strong accents because it doesn’t always understand them.”, “Management by numbers from algorithmic overlords” February 7.

What, should we assume that the capacity of someone in a call center being understood would not be one of the most important factors considered by a human evaluator?

And when O’Connor refers to “the subtle flexibility of human judgment; decisions tempered by empathy or common sense; the simple ability to sort a problem out by sitting down across a table and talking about it.”, I must state that is absolutely not what happens all the time.

Any reasonable algorithm, with access to good historical data, would never ever have concluded, as the human Basel Committee did, that what is perceived as risky is more dangerous to our bank systems than what is ex ante perceived as safe.

PS. Could we envision a world in which more predictable algorithms managed our wives reactions… and, if so, would we then not miss their lovable unpredictability?

@PerKurowski

Our best hope for a decent and affordable adult social care must be minimizing the intermediaries’ takes, whether these are private or public

Sir, Diane Coyle when discussing the possibilities and need for organizing for instance adult social care, and thereto taking advantage of new methods to connect demand and supply and as exemplified by Uber, expresses concern for “the treatment and status of workers in platform public services (although it is not as if these are high-status jobs at present)” “Algorithms can deliver public services, too” December 19.

What’s missing though in that good analysis, is not having contemplating additional tech advances. For example Uber wants to buy self driven cars, in order to get the complications of human drives out of their way, but without realizing that consumers might at one point take direct contact with those cars, in order to get Uber out of the way.

The same will happen for workers in public services, though of course the increased demand for adult social care should help to keep up the demand for many of them. But, even in this case who knows? If you think of yourself as an older person soiled with your own feces, what’s currently is delicate referred to as an “accident”, who would you feel most comfortable with cleaning you, a not too human 1st class robot or a human? 

Sir, the way our generation, and governments have gone on a debt binge, to anticipate current consumption, there will come a time for a reckoning. If we do not find ways to minimize the intermediaries’ take, we will not afford the basic services we need and want.

Of course intermediaries are workers too… and that is why even for them we need to create decent and worthy unemployments.

@PerKurowski

The Basel Committee’s bank regulators being replaced by an algorithm could be the best that could happen.

Sir, I refer to Gillian Tett’s “Self-driving finance could turn into a runaway train”, December 8.

Well human-driven banks are now not doing so well either. 

Any algorithm currently making credit decisions for a bank would do so based on maximizing risk-adjusted returns on equity, based on perceived risks of assets and on regulatory bank capital requirements regulations.

Where would it get the risk perceptions? Currently credit ratings… Who knows if in the future algorithms would also take over the credit rating functions… if these have not already done so?

Where would it get the capital requirements? Currently it get those from the Basel Committee’s standardized risk weights, or if the algorithm works for a sophisticated bank, from its own risk models.

So, if the algorithm does its job well, and works for a sophisticated banks, it would seem that in order to obtain the highest risk adjusted return on equity, its priority has to be creating the risk model that minimizes the capital requirement.

And if it works for a bank that uses the standardized risk weights, then it is clear it would not waste its time with what carries a 100% risk weight, like an entrepreneur, but concentrate entirely on those with much lower risk weights, sovereign 0%, AAA rated 20%, residential mortgages 35%.

So, with the risk weighted capital requirements it is clear that whether the banker is a human or an algorithm, we can forget about savvy loan officers… they will all be equity minimizers.

Of course, an entrepreneur can always offer to pay sufficiently high interest rates to overcome the regulatory handicap. But, would doing so not make him even more risky? With current regulatory risk aversion we should cry for the future real economy of our children.

Sir, in 2003, at the World Bank’s Executive Board (before Nassim Nicholas Taleb had appeared on the scene to discuss fragility) I stated: "A mixture of thousand solutions, many of them inadequate, may lead to a flexible world that can bend with the storms. A world obsessed with Best Practices may calcify its structure and break with any small wind."

So, I guess you can you imagine how much I fret us humans falling into the hands of a final conquering algorithm.

Or having to suffer the consequences of the systemic risks resulting from banks using fewer and fewer human bankers… with probably higher bonuses to the remainders.

By the way since the replaced bankers used to pay taxes, will we at least be able to tax those algorithms?

But, come to think of it, if an algorithm substituted for bank regulators that could be great news. I mean any half-decent algorithm would be able to figure out that what is really risky for our bank system is not what is perceived as risky but what is perceived as safe.

And any half-decent algorithm would also require an answer to the question of “What is the purpose of banks?” And I suppose no regulator would dare tell it, “Only to make the maximum risk adjusted returns on equity” 

@PerKurowski

When being rightly suspicious about making algorithms powerful let us not ignore that powerful humans could be very dangerous too.

Sir, Tim Harford, agreeing with Hayek holds “Market forces remain a more powerful computer than anything made of silicon.” “Algorithms of the world, do not unite!” December 2.

But when regulators decided to replace the risk assessments of thousands of individual and diverse bankers, with those produced by some few human fallible credit rating agencies; and then allowed banks to increase their bets on these ratings being correct, for instance with Basel II allowing banks to leverage a mindboggling 62.5 times if only an AAA or an AA rating was present, we would have benefitted immensely from having some algorithms indicate them this was pure folly.

Because, in the development of such algorithms, it would not been acceptable to look solely at the risks of bank assets as such, but would have required to consider the risk those assets posed for the banks.

And as a result the algorithms would not have allowed banks to leverage more with safe assets than with risky, that because only assets perceived as very safe can lead to the build up of such excessive exposures that they could endanger the whole bank system, were the credit ratings to turn out wrong.

“An Explanatory Note on the Basel II IRB (internal ratings-based) Risk Weight Functions” expresses: “The model [is] portfolio invariant and so the capital required for any given loan does only depend on the risk of that loan and must not depend on the portfolio it is added to.”

And the explicit reason given for that inexplicable simplification was: “Taking into account the actual portfolio composition when determining capital for each loan - as is done in more advanced credit portfolio models - would have been a too complex task for most banks and supervisors alike.”

Sir, algorithms are precisely designed to combat such complexities.

Yes, “Facebook and Google have too much power” but so did the regulators; and with their risk weighting of the sovereign with 0% and citizens with 100%, Stalin would have been very proud of them.

@PerKurowski

Janet Yellen, Mario Draghi, ask IBM’s Watson what algorithms he would feed robobankers, to make these useful and safe

Sir, Sam Fleming reporting from Jackson Hole writes “Janet Yellen, the Federal Reserve chair said regulatory reforms pushed through after the great financial crisis had made the system “substantially safer” and were not weighing on growth or lending. … If the lessons of the last crisis were remembered “we have reason to hope that the financial system and economy will experience fewer crises and recover from any future crisis more quickly”, “Yellen warns opponents of tighter financial rules to remember lessons of crisis” August 26.

As I see it Yellen has not yet learned at all that past and future financial crisis have not, nor will ever, result from excessive exposures to what was or is perceived as risky, these will always result from unexpected events, like when that was perceived, decreed or concocted as very safe, turned out ex post to be very risky.

Since regulators do not to want listen to anything else but their own mutual admiration net-works’ risk biases, I wish they would contract IBM’s neutral Watson to ask it the following:

Watson, while considering the purpose of banks as well as the real dangers to our financial systems, what algorithms would you suggest feeding robobankers with?

THEN Yellen, Draghi and colleagues should compare that algorithm with what they are feeding the human bankers with; the portfolio invariant risk weighted capital requirements that assumes that bankers do not see or clear for risks by means of size of exposure and risk premiums charged.

Then these regulators would understand that with their over-the-board incentives for banks to invest or lend to what is safe, like AAA rated securities and sovereigns, like Greece, they are in fact creating those conditions that dooms banks to suffer huge crises, sooner or later, over and over.

Then these regulators would understand that their regulations induce banks to stay away way too much from lending to what is perceived risky, like SMEs and entrepreneurs, something which clearly must weigh heavy against the prospects of our real economy to growth.

Janet Yellen, Mario Draghi, please ask Watson! Perhaps you could find him on LinkedIn 😆

@PerKurowski

The Basel Committee for Banking Supervision was sold a regulatory algorithm that used the wrong data.

Sir, Richard Waters writes: “a new layer of technology is being added to turn [big data] into the learning that makes applications more intelligent. It represents an emerging tech infrastructure that makes AI (artificial intelligence) not just a new application but a new approach to computing.” “Crowdsourced algorithms promise to be next big thing” March 24.

I register again, for the umpteenth time, the following comment:

Bank regulators, when they defined their risk weighted capital requirements for banks, which so much influences the allocation of bank credit to the real economy, looked at the risk of the assets, and not at the risk those assets posed to the bank system.

As a result they came up with that loony theory that what is perceived as very safe, is safer for the banking system that what is perceived as risky. That’s why they assigned only a 20% risk weight to what can be so dangerous as what is rated AAA to AA, and 150% to the so totally innocuous below BB-rated.

That caused a crisis because of excessive exposures to what was rated AAA; and low growth because of lack of exposures to what is perceived as risky, like to the 100% risk weighted SMEs and entreprenuers.

Sir, that is why all algorithms being developed should be required to carry a warning signed saying “Applying this to the wrong data, can be truly disastrous”

@PerKurowski

If an algorithm can be the boss, why don’t we use our own algorithms to be our own bosses?

Sir, I refer to Sarah O’Connor’s enlightening and interesting “When the boss is an algorithm” September 10.

I would not mind at all getting rid of my car, if I was sure there was a service out there that could respond reasonably well to my needs.

But my needs are in essence somewhat different than Uber drivers’ needs. I want a taxi when I need it, and they offer a taxi when their drivers feel like it.

So, in my neighborhood, and I care little about neighborhoods hundred of miles away, why could we not have a transportation cooperative, run by algorithms decided upon between users and drivers?

In fact, even if I got rid of my own car, I can easily imagine myself providing driving services using my neighbor’s car, with his remunerated permission of course, or using some collective neighborhood cars.

@PerKurowski ©

If only regulators had had one of those “what-to-do” algorithms Tim Harford mentions before regulating banks.

Sir, Tim Harford refers to Brian Christian’s and Tom Griffiths’ “Algorithms to Live By” in order to ask: “Can computer scientists –– help us to solve human problems such as having too many things to do, and not enough time in which to do them? He concludes “It’s an appealing idea to any economist”, among others because “Computers practise ‘interrupt coalescing’, or lumping little tasks together. A shopping list helps to prevent unnecessary return trips to the shop.” “An algorithm for getting through your to-do list” August 6.

How I wish the bank regulators had had access to such algorithms and to the lumping together of all their, not that small, but huge necessary tasks.

If so, they would have been remembered to define the purpose of the banks, among which is the need to allocate credit efficiently to the real economy stands out, and so they would have stayed away from their distortive risk weighted capital requirements.

If so, they would have remembered to read some books on past crises, or looked into some empirical data, and thereby have understood that bank crises are never ever caused by excessive exposures to something perceived as risky, but always from excessive exposures to something perceived as very safe when put on the balance sheet, and so they would have know their risk-weighted capital requirements were 180 degrees off target.

@PerKurowski ©

Who in his right mind can believe credits rated BB- are more risky to the banking system than credits rated AAA to AA?

Sir, on October 22 Gregory Meyer and Joe Rennison reported on FT’s front page “US regulators signal first moves to rein risks of high speed trading”. Timothy Massad, chairman of the Commodity Futures Trading Commission was quoted saying “he wanted to safeguard financial markets against algorithms going haywire” That is great, but who is going to guard financial markets from regulatory algorithms going haywire?

Anyone who dares to really enter and analyze the pillar of current bank regulations, the credit risk weighted capital requirements for banks, comes out not believing what he has seen.

For instance, in Basel II, a credit to the private sector rated AAA to AA was assigned a risk weight of 20 percent while a similar credit to someone rated BB- or less was assigned a 150 percent credit risk weight... meaning a 7.5 times higher capital requirements.

And I just ask, who, in his right mind, can believe credits rated BB- are more risky to the banking system than credits rated AAA to AA? Honestly, what could attract more excessive financial exposures?

@PerKurowski ©