Will AI Predict its Own Future?
A Chat with a Panel of Experts at the IIT Bay Area Leadership Conference 2018
by Devika G. Bansal
Smart homes, all-knowing assistants, self-driving cars — hard to escape these trends if you read the news. And chances are all of these technologies appear to be powered by artificial intelligence, or AI.
Of late, the term has become something of a marketing catchphrase that startups tend to throw around a lot, some more casually than others. But it’s usually not that hard to separate the grain from the chaff, says Raviraj Jain, partner at Lightspeed Ventures. The key, he says, is to look for companies that are not enamored by AI, but are instead working on a problem that is best solved using the technology.
“AI is a hammer, but you can’t assume that everything is a nail,” Jain says. “The question is are you understanding the problem first and starting with that?”
The term “artificial intelligence” first came along in 1955, when a group of scientists proposed a “2 month, 10 man study” to “make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.”
But in recent years, AI has gotten enmeshed in terms like machine learning and deep learning. Simply put, machine learning is a way to engineer AI, and deep learning is just one of the many ways to get machines to learn. Our brains are pattern recognition machines and that’s what these algorithms are designed to do as well — glean patterns from a data deluge.
“To me, the notion of AI is the learning system,” says Rakesh Mathur, CEO and co-founder of WhiteRabbit. “It’s an approach you take because there’s lots of super complex phenomena where variables become super large — and it is not humanly possible to sift through. But a lot of claims that are made are probably way too futuristic. A lot of what happens in the brain is far from being captured.”
Mathur’s stealth startup is using AI to train the largest data set of cancer images to spot cancers when physicians can’t.
“Human beings are very good at identifying predators,” he says. “But cancers look nothing like that. There’s no collective programming that has taken place during evolution for us to be able to recognize that. So that’s an area where you can surpass human ability.”
Another area where machine learning is working well is cyber security, says Jain of Lightspeed Ventures. “Hackers are thinking of different ways to infiltrate all the time, and so you need to do very smart pattern recognition to prevent attacks.”
Banks have been using the technology in fraud analytics for quite some time now, says Debu Chatterjee, AI engineering leader at ServiceNow. Hedge funds that provide high level financial services are also folding in AI because investors need to make quick decisions based on a gazillion signals.
So what’s the next big problem AI could solve? It could be to have a meaningful conversation with your virtual assistant, or a quick resolution of your service desk ticket, says Chatterjee. At the other end of the spectrum is healthcare: the idea of personalized medicine and the ability to predict illnesses or heart attacks, says Jain.
“Eventually, companies that are using AI as a hype will probably disappear,” adds Chatterjee.
But where we’re at right now, perhaps only AI can predict how far the technology might take us into the future.