Why AI will not replace experts

Sonja Georgievska
Netherlands eScience Center
3 min readMar 3, 2020

Forget about “responsible AI” for a moment. Data, not AI, is the product, and it needs rigorous quality control.

You have probably heard the story of a neural network that learned to recognize images of horses by “looking” for a copyright tag that happened to be associated with horse pictures, rather than learning to recognize a horse in the picture. Or a similar story. Many stories such as this one have led to a bad reputation for so-called Artificial Intelligence, or AI, and initiated the need for Explainable AI, Accountable AI, Fair AI, you name it. “AI is a black box that transforms innocent data into a product, but the product cannot be trusted”. “We need to open up the black box, or else the society is doomed”… (?!)

Let’s take a few steps back. It is 1990 and a team of medical researchers equipped with a statistician is performing a randomized control study to find out if phenomenon X causes disease Y. They first take the utmost care that the data, I mean the patients, from which they want to derive conclusions, are representative of the population; that the experimental and the control group are properly established; that there are no correlations in the data-set, sorry, cohort, that might contaminate the results of the study; etc, etc… You get the point (I am not a professional statistician, so please ask one for more details). Only after the experiment is prepared with the utmost rigor, are they allowed to actually make the experiment and apply AI, I mean, statistical methods, to gain predictions. (Did I say predictions? I meant conclusions.)

“We don’t know why, but a state-of-art algorithm told us you can take these”. Photo by Kendal on Unsplash

It is now 2020. Imagine that a team of medical researchers collects a huge amount of profiles of “patients” from a popular social media website, and says “Bingo! We have so much data! Let us apply one of the fancy new statistical, now called AI, methods, and get results without all the effort we used to put into it”. They get the results, make a pill based on them, send it for mass production, the pills start killing instead of curing… It is difficult to imagine, isn’t it? Then why on Earth do we want the prediction methods of the 21st century to be able to turn garbage data into gold? Is it because we gave them such a sophisticated name, Artificial Intelligence, and so we expect them to replace humans altogether?

How about this: instead of saying that AI turns data into prediction, say that it merely consumes the data to be able to perform its daily activities. (So, don’t worry, we are still humanizing AI ’cause it’s so irresistible). So, data is food for AI. When we buy food from the supermarket we know that it has gone through quality control during production to make sure we don’t eat poison. When we pick up spinach from the garden we wash it thoroughly to make sure it is not contaminated with bacteria. Or else we get sick and anything can happen. But from AI we expect it to be almighty and be able to skip all those quality control steps, ’cause AI is, after all, supernatural and such creatures cannot get sick, can they?

So, here is the good news. AI is not going to replace the jobs of radiologists, medical experts, accountants, please finish the list. It is just going to transform the kind of jobs that they are doing and speed up progress. Nobody is going to make AI almighty. Experts will be needed to make sure AI is being fed with the right data. Only then will society be able to trust AI. Because it will not be about AI in the end. It will be about the people, again.

Thanks to Patrick Bos, Florian Huber, Carlos Martinez Ortiz, Tom Bakker, and Johan Ridder for the useful remarks.

Sign up to discover human stories that deepen your understanding of the world.

Free

Distraction-free reading. No ads.

Organize your knowledge with lists and highlights.

Tell your story. Find your audience.

Membership

Read member-only stories

Support writers you read most

Earn money for your writing

Listen to audio narrations

Read offline with the Medium app

Published in Netherlands eScience Center

We’re an independent foundation with 80+ passionate people working together in the Netherlands’ national centre for academic research software.

Written by Sonja Georgievska

“Insides out” of the AI bubble v. 2010s-’20s

No responses yet

What are your thoughts?