Machine learning is the tail that wags the AI dog.
Sure it looks as if artificial intelligence is dominating the news. In fact, we are awash in AI hype that tends to fall into seven basic categories:
- AI Gains Astonishing New Skill Sets: There’s a constant flow of news about how some AI has acquired a new skill that had, up till then, been the strict purview of humans. For example, there have been reports about about how AI is just as good as human beings in everything from playing the game of Go or Jeopardy to detecting breast cancer. There are also AIs that can paint, write songs, help the blind recognize objects, come up with near cancer treatment ideas, make medical diagnoses and treatment plans, write news articles, etc.
- AI Will Steal Your Job: As they skill up, of course, there’s a worry that AIs will take over more and more jobs. The potential list seems to get longer every week: stockroom worker, bartender, pharmacist, journalist, housekeeper, paralegal, truck driver, cashier, surgeon, and many more.
- AI Apocalypse (or Rapture) Is Coming: Over the last year or so, some heavy intellectual hitters have warned the world about the dangers of strong AI. They include people such as renown physicist Stephen Hawking, Microsoft co-founder Bill Gates, Apple co-founder Steve Wozniak, and business magnate Elon Musk. In fact, even some Google researchers are worried enough about out-of-control AIs that they’ve built a “kill switch” (though they may be worried more about screwy housekeeping bots than the Terminator). On the other side of the fence, people such as computer scientist and futurist Ray Kurzweil believes that The Singularity–that is, the point at which AI surpasses human intelligence–will bring about an era when human beings enjoy immortality and astonishing powers. Some pundits refer to this, typically tongue-in-cheek, as the “rapture of the nerds.”
- AI Isn’t There Yet: The fourth type of prevalent AI article tends to quote computer scientists and others who argue that we’re still far from developing a truly strong AI (despite Kurzeil’s claim that The Singularity is Near). We typically get a rash of reassurances just after an AI badly humiliates some expert human in a game that, until then, he or she had dominated. Think chess, Jeopardy, Go and even rock-paper-scissors. An example is a recent New York Times article wherein an obligatory panel of scientists comforted everyone, saying that AIs are still far from matching overall human intelligence (good to know).
- Business Investments in AI Are Staggering: Maybe we’re headed toward robopocalypse, but there’s plenty of dough to made in the meantime. The Motley Fool, for example, just published stats on AI investments. Among them:
- $5.4 billion invested in AI start-ups
- A $5.05 billion market by 2020
- 6 billion devices will request AI support
- AI Will Augment Us: If you can’t beat them, join them. That’s the theory behind using AI to augment human capabilities. Neil Jacobstein, who chairs the Artificial Intelligence and Robotics Track at Singularity University, believes business organizations can use forms of AI to boost the pattern recognition abilities of employers: “If the organization has committed itself to evidence-based decision making, and they use AI for pattern recognition, they can do, I think, a much better job in avoiding bias and in paying attention to patterns in the world that humans unaided would have difficulty seeing.” Further down the road, he envisions, AIs will be able to emulate the human neocortex and so extend our problem-solving abilities.
- The future of AI: And then there are the articles that look ahead at AI in general. They typically incorporate some mix of the ideas discussed above, but they focus on the question of “what’s next in the AI world?”
Okay, so back to the topic of machine learning. Yann LeCun, director of AI research at Facebook, notes that it is machine learning processes that allow AIs to improve in any given area. “Show millions of cat photos to a machine, and it will hone its algorithms to improve at recognizing pictures of cats,” he writes. “Machine learning is the basis on which all large Internet companies are built, enabling them to rank responses to a search query, give suggestions and select the most relevant content for a given user.”
In essence, machine learning is the secret sauce that permits AI algorithms to become more intelligent than a programmer could otherwise make them. It tends to be the foundation of an AI’s acquisition of new skills, its ability to take on yet another human job, and its capacity to recognize patterns so well that it becomes the target of billions of dollars worth of investments.
The power of machine learning helps explain the growing usage of the phrase both in professional circles and the larger popular culture. This can be seen in the following graphic from Google Trends:
The Growing Popularity of the Term ‘Machine Learning’
It’s no mistake, of course, that machine learning methods are strongly related to predictive analytics techniques. Prediction often relies on pattern detection, which may be the single greatest strength of machine learning. This raises all kind of questions around how prediction, cognition, and learning are related, but that’s a subject for the future.