Sir, Murad Ahmed, writing about Demis Hassibis states: “At DeepMind, engineers have created programs based on neural networks, modeled on the human brain. These systems make mistakes, but learn and improve over time” “
Master of the new machine age” March 12.
Ooops! I hope they do not use as models the brains of current bank regulators.
In 2007-08 we had a big crisis because AAA rated securities and sovereigns like Greece, perceived and deemed as safe, turned out to be very risky.
And what connected all that failure, was the fact that banks were allowed to hold very little, I mean very little, we are talking about 1.6 percent or less in capital, against those assets, only because these were ex ante perceived or deemed to be very safe.
Of course, anyone who knew anything about the history of financial crises would have alerted the regulators that to allow banks to have less capital against what is perceived as safe than against what is perceived as risky, was very dumb. That since major crises only result from excessive exposures to something ex ante perceived as risky but that ex post turns out to be very risky. And one of the main reasons for that is precisely that too many go looking for “safety”.
But now we are in 2016, and the issue of the distortion those capital requirements produce in the allocation of bank credit to the real economy is not yet even discussed.
So before these human brain systems learn and improve over time from mistakes, they have to be able to understand these and, more importantly, to humbly accept these.
Frankly, artificial intelligence seems it could have an advantage over humans’, namely none of that human ego that so much stands in the way of admitting mistakes.
But also beware, were robots free of that weakening ego, they could conquer us!
@PerKurowski ©
Posts
