I read a Pew Research Center online publication over the last few days. A large number of people, many of them experts in computer science, social science, anthropology, directors or researchers at think tanks, doctoral candidates and professors, writers, and others (a few) not specified, were asked to answer specific questions about AI. I read every comment over several pages. They were writing on the topic of AI and its likely social effects ongoing and in the future for about 13 years until 2030.
To learn more about U.S. politics, economics and history, buy a copy of Political Catsup with Economy Fries, available at Amazon.com.
If you want to visit this online publication, here’s the link: https://www.pewresearch.org/internet/2018/12/10/concerns-about-human-agency-evolution-and-survival. I would like to take a moment to thank the respondents and Pew for providing this written discussion to me and you.
The AI that currently exists isn’t really artificial intelligence like a person might be intelligent. Instead there are narrow applications for computer learning and computer moderated data processing. This narrow AI is unlike broad general intelligence. It is broad computer intelligence, fortunately not existing yet, that makes people the most fearful. That is what scared Stephen Hawking during his life. He warned that true general level artificial intelligence in a computer system could cause the end of human society.
We see changes all around us because of narrow applications of AI that are constantly altering the way human energy is channelled into human environments these days, whether the work environment, the home environment, education and other forms of training and learning and even marketing and entertainment. We are inside a changing environment that is being altered by AI algorithms.
There’s almost no accountability for the operation of these algorithms. If the AI algorithm caused you harm, you have no recourse even to discover how it harmed you. You can’t ask anyone to stop the algorithm or even to change it because no one is available to talk to you about it. This is causing tremendous pain and we have seen pain like this in the recent past but in chemistry instead of computer science.
An unfortunate thing about the idea of “better life through chemistry,” has been our historical willingness to assume that all new chemicals are harmless when a new chemical is created by chemists. There are numerous examples of chemicals that have caused harm to people’s health and to the environment. Only after much damage has been noticed and carefully traced to the chemical at fault is the harm finally limited by restrictions on the harmful chemical.
By the time the bad chemical is no longer sold, the company that created it has already profited and much of the harms caused by the chemical will not be paid by the producer. Instead, health problems will be absorbed by those exposed and harmed. The environment will need time to slowly recover from harms. And there’s no telling whether people or the environment will be able to carry on very well even after the harmful chemical is removed.
Sometimes huge expenditures are necessary to help those that are harmed from eagles to children. DDT is the one of the most ready examples that comes to mind but so are fluorocarbons like the ones that were in ScotchGuard and that are now known to endanger health even in tiny quantities and also unfortunately to persist for a long time in our environment.
We’re treating algorithms like they are harmless as a matter of course when we know that some algorithms create social harms and individual harm.
What surprised me about the experts in the Pew query, the people who know AI the best and that interact in the environment of discussions about AI was twofold.
First, I noticed that the Pew group of experts were mostly providing negative judgements about likely harms that they expect from AI that will harm some individuals in every society. The pervasive negativity of their judgements was daunting to me. Secondly, these leaders in understanding AI acted mostly like powerless observers who felt unable to change the flow of harms in the context of today’s and tomorrow’s algorithms toward a more positive outcome for more people.
I’m reminded of the Genesis song, “Land of Confusion,” words and music by Tony Banks, Phil Collins and Mike Rutherford from the 1986 Invisible Touch CD/album. The lyric I have in mind is,
“Ooh Superman where are you now
When everything’s gone wrong somehow
The men of steel, the men of power
Are losing control by the hour”
In the shambles being caused around us by algorithms, we all have to be like Superman to right the wrongs that are tearing apart our society. That means that algorithms can’t be assumed to be harmless. A lack of accountability to harms is itself harmful. Lack of accountability makes wrongs more pervasive and long lasting.
Instead of thinking that we are all helpless, let’s assume that we have the power to stop harming our societies by paying attention to harms being caused by algorithms. We can take steps in our society’s politics through legislation to right algorithmic wrongs. We can start by making algorithm creators accountable and people who use algorithms accountable for harms that they cause to others. We don’t have to remain helpless in the face of this technological change.
Harms caused by AI continue in the meantime. Examples are increasing. In my community, I’m seeing help wanted signs outside businesses. I think the reason these job advertisements are appearing is that people have little confidence in online job services. One employer told me that these services don’t produce the promised perfect candidate. Others have found that some of the online job services have poor security.
Another example is 35 cycle testing PCR for virus. The inventor of PCR technology said that PCR can’t be used to detect a virus accurately: it will produce false positives (even with a lower threshold than 35 cycles). The idea that we must continue with what isn’t working is an example of letting an algorithm linked to a faulty PCR test produce false positives without end–as though that policy decision never needs to be evaluated.
Both of these examples show a failure of accountability and a failure to fix a problem. This failure is costly in many ways and it should be addressed.
Finally, whenever an expert gives you their opinion that humankind may be doomed by technologies as they unfold, don’t believe them when they bemoan this as an inevitability. We can choose a better future for ourselves and for our society.