The Limits of Artificial Intelligence

There are 10 key technologies that will drive the Next Economy, the one that underpins them all is artificial intelligence. And while AI is already delivering value, many challenges are left to be overcome before it truly reaches its potential.

As someone who’s developed emotion recognition technology and have talked to leaders, organizations and groups about the limits of artificial intelligence, I was delighted to read a contrarian perspective on artificial intelligence on the NY Times: Gary Marcus argues that AI must account for basic concepts of how the world works, like time, space, and causality, beyond statistical pattern detection, before it can earn our trust.

Before we can expect it to have some sense of common sense, the main challenge with artificial intelligence is bias and explainability. So, in order to trust artificial intelligence we must start over with first principles like time, place and causality.

Here are a few quotes from the article:

“Without the concepts of time, space and causality, much of common sense is impossible.”

“We need to shift our approach to artificial intelligence in the hope of developing machines that have rich enough conceptual understanding of the world that we need not fear for their operation.”

“Dystopian speculation arises in large part from thinking about today’s mindless artificial intelligence systems and extrapolating from them. If all you can calculate is statistical correlation, you can’t conceptualize harm.”

The article is worth reading, let me know what you think!