What are the actual risks to society associated with the widespread use of AI? Outside of the use of AI in a military context.

I am not talking about accidental risks or unintentional behaviour - eg, a driver-less car accidentally crashing.

And I am not talking about any transitional effects when we see the use of AI being widespread and popular. For instance I have heard that the widespread use of AI will make many existing jobs redundant, putting many people out of work. However this is true of any major leap forward in technology (for example the motor car killed off the stable/farrier industries). The leaps forward in technology almost always end up creating more jobs than were lost in the long run.

I am interested in long term risks and adverse effects stemming directly from the widespread use of AI in a non-military sense. Has anybody speculated on the social or psychological impacts that AI will produce once it has become popular?


The biggest risk is algorithmic bias. As more and more decision-making processes are taken on by AI systems, there will be an abdication of responsibility to the computer; people in charge will simply claim the computer did it, and they cannot change it.

The real problem is that training data for machine learning often contains bias, which is usually ignored or not recognised. There was a story on BBC Radio about someone whose passport photo was rejected by an algorithm because he supposedly had his mouth open. However, he belonged to an ethnic group which has larger lips than Caucasian whites, but the machine could not cope with that.

There is a whole raft of examples where similar things happen: if you belong to a minority group, machine learning can lead to you being excluded, just because the algorithms will have been trained on training data that was too restricted.

Update: Here is a link to a BBC News story about the example I mentioned.

  • $\begingroup$ What about deepfakes and it's unsavory variants? $\endgroup$
    – user9947
    Oct 24 '19 at 9:06
  • $\begingroup$ @DuttaA Yes, don't get me started on that! However, I think algorithmic bias is a bigger risk, as there is little awareness of the issue. People can learn not to trust videos (as they cannot trust texts) alone, but with bias there's nothing you can do. $\endgroup$ Oct 24 '19 at 9:09

One risk that’s already realized: large online vendors think they have implemented artificial intelligence in their “help” pages and therefore they can (try to) make it impossible to get to someone who can actually think. And since the artificial stupidity (AS) usually feeds the customer articles completely unrelated to the issue, anyone sufficiently persistent to pursue it is extremely pissed off at the company before (if ever) the issue is resolved. And because far too many people passively accept this abuse, the companies have no incentive to be more reasonable. In other words, “AS” is reducing our expectations for customer service.

Another is the JavaScript intended to prevent invalid names, phone numbers, and email addresses in web forms which due to bugs or obsolescence rejects legitimate inputs.


IMHO the greatest risk is that AI can make people lazy. If you can ask an AI for an answer to any problem, what's your motivation to figure out how to figure out the answer for yourself? I have run into a lot of young people who can't add or multiply two three-digit numbers without using a calculator. When it's possible to dump a huge mass of data into an AI, and the AI tells you the structures it finds in the data without explaining how it finds the structures so you can do it yourself, the AI wins and you lose.

  • $\begingroup$ when making a fire, do you use sticks to create friction to burn some kindling - or do you use a lighter? when travelling a long distance, do you walk, ride a horse - or do you drive a car or take a train? when you want to write something down, do you make impressions into clay, or do you use a pen and paper? .......and does the fact that you don't rub sticks together, or walk long distances, or make impressions into clay mean that you are lazy? $\endgroup$
    – Jimmery
    Oct 26 '19 at 16:53
  • $\begingroup$ The Wall-E scenario! $\endgroup$
    – DukeZhou
    Oct 26 '19 at 20:11
  • $\begingroup$ @Jimmery, that's a reasonable argument. However, human beings design and make the lighters, cars, pens, and paper. And we do remember how to walk, make impressions in clay, etc. But how to think and how to create are cultural and easily lost. If we rely on AI to generate our art, we cease to be artists. If only AI "understands" cosmology, we cease to be cosmologists. If only AI does sports, we will only be spectators. AI is extremely useful, but we should use it to strengthen humankind, not let it make us mere consumers and spectators. $\endgroup$
    – S. McGrew
    Oct 28 '19 at 12:24
  • Offloading of responsibility may the single greatest danger.

Where algorithmic bias may be the core issue of Machine Learning, it can be identified and mitigated.

Transferring responsibility to a robot or algorithm requires an intentional choice with moral dimension. As the scholar Joanna Bryson put it:

In humans consciousness and ethics are associated with our morality, but that is because of our evolutionary and cultural history. In artefacts, moral obligation is not tied by either logical or mechanical necessity to awareness or feelings. This is one of the reasons we shouldn't make AI responsible: we can't punish it in a meaningful way.
Source: AI Ethics: Artificial Intelligence, Robots, and Society

In a malicious sense, transferring agency to an automaton that may do something harmful which benefits me allows me to say "I didn't make the decision and have no responsibility for the outcome." (It seems to me that companies are doing this more and more.)

There was a very good short story on the subject Unchained: A story of love, loss, and blockchain in which automate taxis develop novel strategies that have an unintended moral dimension in regard to humans.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.