Does A.I really pose a threat to humanity?

The term artificial intelligence has been around since 1956 and is not precisely defined. Whatever you mean by it, it poses no threat to the global existence of humanity. You’ve been taking too seriously the utterances of a few prominent people who should know better.

What we should be taking seriously are the actual threats from a specific type of AI, machine learning. This technology has found a very wide range of applications and its potential is virtually unlimited. However, it has some known problems and there may be more as it works its way into every part of society. These issues tend to be the most discussed by AI researchers and the popular press:

Future unemployment - the most hyped threat, actual effects are difficult to predict;

Not 100% accurate - Usually as good as human performance, which is not a high bar;

Training data can be biased - an aspect that’s difficult to assess and correct;

Errors are not traceable - deep neural networks are a logical rat’s nest;

Not transparent to users - Was that decision made by a machine or a human?

Can be hacked - malicious experts can influence results;

Misguided applications - autonomous weapons, smart viruses, deep fake videos, speech and text, plus many more existing and potential unethical uses.

Machine learning is a whole new kind of technology and it can be tremendously beneficial to humanity. It can also be both intentionally and unintentionally harmful. The European Union has issued guidelines for AI and automated decisions, which may eventually become regulations. The rest of the world needs to follow suit.