Most people aren't aware of the latest advances in artificial intelligence or don't realize the substantial progress that's been made in recent years. How far will this progress expand, what are the different approaches currently being used, and is something like superintelligence possible? These are some of the questions we recently explored on our podcast with Pedro Domingos on his latest book, The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World.
What Is Machine Learning?
Machine learning is a subfield of artificial intelligence, and the goal is to build computers to do the things that traditionally require human intelligence, Domingos stated. This can range from seeing, understanding language and reasoning, to basic common sense and pattern recognition.
“The goal of machine learning is to automate learning,” Domingos said. “It is the key field in AI, because it’s the one that enables all the others.”
Machine learning, in turn, draws its power from big data, which is the reason we’re seeing a lot of progress in this area.
“One way to think of this is, AI is the planet we’re headed to, machine learning is the rocket getting us there, and big data is the fuel for that rocket,” Domingos said.
Here's a short 3-minute clip from our recent podcast interview to hear what he had to say:
Deep Learning Dominating Currently
Deep learning, which is actually a type of multi-layered neural network modeled after the human brain, has scored a lot of successes in the last 5 years or so, Domingos said, adding that it’s definitely the most popular approach right now.
“I think this is what a lot of people miss,” he said. “When they think of AI, they think of agents with separate goals from us. Why would we create those? AI is really an extension of our intelligence. It’s actually something that makes us more powerful.”
A functioning AI would allow us to search more, look deeper, and in essence become more powerful. Often, the worry of these people is that computers will get too smart and take over the world, Domingos stated, but the real danger is computers that aren’t smart enough. Computers already make a lot of decisions on our behalf, he noted, and we’re better off if they’re more — not less — intelligent.
Super Intelligence Inevitable?
Domingos believes we are moving toward a Grand Unified Theory of machine learning. This may render a “master algorithm” that can solve any problem if we give it the appropriate data to learn from.
The biggest misconception is the optimistic notion that AI is easier to achieve than it seems, Domingos added. The second misconception is that human beings are somehow magical, and computers will never reach our levels of intelligence.
“I don’t think that’s the case,” Domingos said. “I don’t think there’s a fundamental limit on how intelligent computers can be, other than the fundamental limit on what you can predict or do by any means.”
Listen to this full interview with Pedro Domingos on his book, The Master Algorithm, by logging in and clicking here. Become a subscriber and gain full access to our premium podcast interviews with various guest experts by clicking here.