Don’t worry, Artificial Intelligence May Never Be Enough…

Original article was published on Artificial Intelligence on Medium

Don’t worry, Artificial Intelligence May Never Be Enough…

Okay, I may have a some good news, AI May Never takeover your job.

We are living the era of a re-emerging artificial intelligence or what some people call an information revolution war where machine learning, deep learning and buzz words of the like are the weapons of choice for this war.

But, sorry AI may actually never be enough ever

In theory, with enough data artificial intelligence can learn almost anything, flight an airplane, cook a meal, drive a car, or give haircut, diagnose a cancer, plenty of research paper on how to solve this and that very complex problem is available out there, and they even make the source code available.

But in the reality, this is true only for very basic tasks. The tasks that any human being can do, in fact, it can only do the boring and repetitive tasks that humans do want to do anyways. A machine can only be programmed and trained to do it faster and cheaper, but it will never do tasks that require complex thinking. We must admit that AI applications perform real good at recognizing cats and dogs in pictures !

I recently heard the CEO of a FANG company announcing that they have and will eventually train an artificial intelligence algorithm to replace all human verification that they have to do on reported harmful or illegal content. I will argue that this will never be enough and 9 time out of 10, the ban will be appealed and will need a human to review it, simply because as intelligent as it can be, an AI is a simple stupid algorithm: a set of random weights, fed with data and probabilities, statistics and approximations taken to a higher level due to the very high computation capabilities.

I will also argue that, faster and cheaper is not the best alternative even to execute boring, repetitive tasks. Fast and cheap does not guarantee good quality, and an algorithm cannot be held responsible for a harmful decision just because the data “suggested” that decision.

Originally published at on May 16, 2020.