Dr. Rand Hindi says that without emotional intelligence machines will never be able to obtain human-like artificial intelligence. Reminiscent of Data on Star Trek Next Generation, Hindi says that despite the impressive ability of machines to learn from other machines and to solve logic problems better than humans, most decisions humans make are actually emotionally driven and machines simply don’t have an emotional IQ.
Dr. Rand Hindi, CEO of cutting edge AI technology company Snips, recently talked about the future of AI at LinkedIn Talent Connect:
Human-Like AI is Never Going to Exist
I want to talk to you about the reason why I believe that human-like artificial intelligence is never going to exist and what that means for the future of work. Artificial intelligence is the ability to reproduce human behavior in a machine. That’s it. Take what a human can do intelligently, put it in a machine and you’ve got artificial intelligence.
Within AI you’ve got one type of way to achieve this which is called machine learning. The idea of machine learning is that you’re effectively teaching machines to reproduce the behavior by giving it examples. It’s a little bit like a kids book where you have pictures of animals and the name of the animal is written and then after seeing a few pictures of horses your kids know how to recognize horses. It’s exactly the same thing in machines.
Machine Learning is a Very Big Deal
Machine learning is a very big deal because up until now when you wanted to automate something a human had to first understand what was going on, then sit down and program a machine to do that. Automation was limited to what humans were able to understand. With machine learning all you need is data collected from what you’re trying to automate and the machine does everything else. You no longer need a human expert in the loop.
Within machine learning there is one type of algorithm that’s called deep learning. Deep learning is a branch of machine learning which is a branch of artificial intelligence and you could consider all three to be interchangeable today, but that’s going to change in the future. I don’t believe that word artificial intelligence is actually going to be used in marketing in the next few years.
Deep Learning Has Been a Huge Revolution
How have we been using deep learning? Deep learning has been a huge revolution. We see it happening for self-driving cars. We see it happening for medicine. Medicine is a very important use case for artificial intelligence because we have today AI that can diagnose
You’ve probably heard about Alexa, the voice assistant from Amazon. This is one of the fastest growing consumer products ever. Rumors are that one in six Americans uses that. This was not possible before because deep learning was not enabling us to talk to machines as naturally speaking.
Is AI Getting Out of Control?
Let me tell you about the world champion playing against Google’s artificial intelligence at the game of Go. The game of Go was considered to be extremely complicated for an AI to beat because the number of different combinations meant that the only way for a machine to beat a human is to actually learn how to play the game. We thought this was still ten years in the future. The way that they made this work was really interesting.
They took one artificial intelligence and they made it play against another one. So one was playing the white side, one was playing the black side, but the trick is that every time one of the AI played a move the other one gave it feedback on that move. By mutually reinforcing each other by playing millions and millions of games, eventually they learned how to play the game better than any human.
At the time, when they played against a world champion the world champion won one out of five games so this was amazing. But people felt a little bit reassured, they were like a 20 percent chance of surviving AI that’s still not bad! However, the same AI kept on learning. Today, not a single human can beat that AI at a single game. But there is more, there is a new version of this AI that beats that AI that beats every human at every game.
When I saw that I was like, oh my god, this is just getting out of control, this is getting out of our hands. But what you need to understand is that everything I just talked about, however impressive it is, is still something that is called narrow artificial intelligence. Effectively, those machines are able to do one thing, perhaps do it better than a human, but they’re only doing this one specific thing. The AI that played the game of Go that was a breakthrough, but it doesn’t know how to do anything but play the game of Go.
Machines Will Never Have Emotional Intelligence
Now people are working on something that’s called general artificial intelligence. This idea that a machine could solve any logical task, that it could reason, that it could transfer the learning it had from something to something else. This is major because if you can have a general form of intelligence and reasoning then potentially machines could do anything that seems like intelligence.
But this is still not what you see in movies. What you see in movies is a very human-like artificial intelligence. It’s not just reasoning and logic, it also includes the ability to emotionally connect with humans. So artificial human intelligence is really this combination of logical intelligence and emotional intelligence. It’s IQ plus EQ. If you only take IQ into the equation you don’t end up having human intelligence.
Why is emotional intelligence important? It’s a way that as humans we can solve paradoxes. A paradox is a mathematical problem for which there is no logical solution. If a machine is not able to have emotional intelligence it will never be able to solve logical traps, which means as humans we can use our EQ to find traps for machines that have very high IQ.
You might be thinking that machines are building emotional intelligence as well. We see all those amazing robots, people develop feelings for those robots as well. However, I believe that you will never have true emotional intelligence in machines.
EQ First Requires Artificial Consciousness.
Emotional intelligence first requires artificial consciousness. They will also need to feel emotions. This is very different than pretending to have emotions. It’s very easy for me to learn that when someone is smiling that person is probably happy and it’s very easy for me to smile back. But hey, I can be smiling but it doesn’t mean I feel happy right now. There’s a big difference between perception of emotion, between display of emotion and feeling emotions.
We know that humans who don’t feel emotions are incapable of making decisions on a daily basis. They can do math, they can solve mathematical puzzles, so they have very high IQ potentially. But if you ask them what they would like for lunch they cannot answer because there is no algorithm to answer that question. Given that as humans most of our decisions are emotionally driven. Let’s be honest, we use the data to back it up but we make emotional decisions mostly. A machine that doesn’t have an emotional intelligence will never be seen as a human-like type of intelligence.
What I’m trying to get to here is that yes, you will have an AI that has an IQ of five billion and yes every logical task is potentially doable by a machine, but humans will have the monopoly on emotional intelligence. Humans alone will be able to do emotionally driven tasks and so rather than think about machines replacing humans we really have to start thinking about humans and machines working together.
How can we leverage the horizontal emotional intelligence of humans with the powerful mechanical logical intelligence of machines? Rather than try to build an AI that replaces humans completely, why don’t we start building an AI that actually works with a human in a very natural and very intuitive way.