Algorithms don’t think. They just execute. A lot of jobs/services/functions that were done by humans are becoming algorithms. My job as a floor trader was basically eliminated because algos took over. There will be a lot of jobs eliminated in the future because algorithms can replace them and make the whole process more efficient.
If you can distil a job into an if/then decision and it’s a commoditized action, you can plug an algorithm into it.
Algorithms are not true artificial intelligence. They execute. They don’t think for themselves and learn.
There is a lot fear around algorithms because they are unknown. Except, we have been interacting with them for years. When you go through an application process online, it’s via an algorithm. A decision tree was set up and programmed. If you went to engineering school you most certainly were exposed to algorithms. If you went to business school hopefully you were too. Even back in the dark ages when I was in undergraduate business school we were exposed to decision trees and algorithms in Operations Research.
What’s interesting is that now, people are saying there is bias in algorithms. I am not as sure as Nate Silver is about that.
I don’t think being unregulated is a problem. Algorithms work on data. Input crappy data you get crappy data out. I don’t think we need the government to step in. Ronald Reagan was right when he said the worst words someone could hear was “I am from the government and I am here to help.”
Silver and others are trying to make the case that many algorithms used by public law enforcement are racially biased. The designers of the algorithms are rebutting those assertions.
Without delving deep into the data they are using and understanding how the algorithm is put together it is pretty hard to have a solid opinion on it no matter which side you might have a positive cognitive bias for.
I recall the CFTC wanted access to the proprietary algorithms of high-frequency trading firms. It was totally stupid. They never wanted access to trading strategies before when it was human. When it comes to markets, algos are tweaked all the time.
We are still working through a lot of the nits and knots of algorithms in public markets. I have seen things happen that never happened before. Flash rallies and crashes. Liquidity there then drying up. Algos have broad based adoption but are still in their infancy. There are things in market structures that can be set up to make playing fields level. I think the same goes for other industries and government.
When it comes to industry, the principle of competition has to be paramount.
The concepts Silver writes about in his blog post are all subjective and conditional. That goes against the grain of positive economics and is at the core of a lot of debate we have currently in America regarding all kinds of topics. The problem people have with positive economics is the outcomes they desire aren’t generated. So, they have to “fix” it. Fixing it often leads to more unseen problems than not fixing it. When you make decisions and adopt policies based on positive economic principles, it’s generally more efficient, flatter and fair for everyone.
For many of the decisions that Silver is outlining, algorithms should be a tool, not a be all end all. For example, do mandatory sentencing guidelines hurt minorities more than algorithms? Or should a judge have leeway?
I would agree that when it comes to public policy, algorithms should be transparent to a point. If a private contractor is selling into a government agency there has to be some secret sauce left behind the curtain. Otherwise, there is no competitive advantage. When stuff gets into pure government hands it never works or is so clunky it’s impossible to navigate through. As a society, we want to incentivize innovation. Taking profit motives away will kill innovation.
I also know this. If you start from the premise that you are going to use algorithms to fix human society since human society is imperfect, you will fail. Humans are imperfect. The Utopia a lot of people desire is only in one place, the afterlife.