It's a great time to be a startup in artificial intelligence. It's also a hard time to be a startup in artificial intelligence.
So it goes with startups in hot tech markets all the time, but AI is especially hot these days, thanks to big breakthroughs that have been applied to improve speech and image recognition. That's making for lots of new opportunities, and lots of competition for those opportunities
Not surprisingly, a panel of AI startups (plus one somewhat older company called
Besides the ready availability of capital, one upside for AI startups is that they have access to huge pools of computing power (such as Sentient's network) they never did before. As a result, startups can credibly attack real problems, and not necessarily have to create their own Google-scale networks of computers. "You can solve not just the cartoon version of a problem, but the real version of the problem," said Nigel Duffy, chief technology officer at Sentient Technologies.
And there are plenty of problems to solve, otherwise known as opportunities to exploit. Many go well beyond well-known areas such as speech and image recognition or even robots and autonomous cars such as Google's. Lauri Saft, VP of IBM's Watson Ecosystem & Partner Program, said IBM expected mainstream applications such as banking would be the obvious target for Watson's AI work, but quickly found interest in fields as diverse as veterinary medicine and fantasy sports.
Another potent opportunity could be software. "A lot of what we build is held together by handwritten software," said Adam Cheyer, cofounder and VP of engineering at Viv. "AI can play a huge role in becoming more assistive in handling the problems of creating software."
There's also a rush to provide intelligent assistants, such as the one Facebook announced yesterday--and the one that Viv, still in stealth-ish mode, looks to be building. Better natural language understanding will be a big part of what enables those, said Kevin Quennesson, an engineering manager and staff engineer at Twitter Cortex, the company's AI group. And the list of AI-driven business opportunities for startups goes on: drug discovery, e-commerce, education.
The downside of the current gold rush? Mainly competition, especially for talent. And especially from big companies with virtually unlimited money, such as Google and Facebook that are vacuuming up AI scientists like crazy. The entrepreneurs tried valiantly to point out that there are advantages to working at a startup, such as the potential to work on problems too far-out for established companies (though Google certainly works on its share of those). "Breakthroughs are best done at a startup," said Cheyer. "At a startup, you can count the number of days before you die," which provides a "unique focus." But Norman Winarsky, an advisor to SRI's venture arm (which spawned Siri, later bought by Apple), pointed out the obvious: "It's hard to recruit people from great companies."
The biggest challenge for AI startups is resisting the urge to take what is after all a set of technologies with very broad applications too broad. The key, said Winarsky, is focusing on single opportunities first where AI technology is uniquely applicable, rather than trying to create a broad AI platform. The latter is more likely to be dominated by huge companies such as Google and IBM, he said: "If you're going to climb Mount Everest, you gotta have a base camp first."
Of course, there's an upside to competition as well for startups. Those that are successful can sell out to the big guys for a whole lot of money. That possibility may be driving venture investment as much as the core opportunities, but again, that's how the tech startup world works.
By the way, these folks aren't too worried about the two constant elephants in the AI room: the potential for AI to eliminate a lot of people's jobs, and the potential for AI to eliminate people altogether. On the job front, they think AI will mostly serve to augment humans rather than replacing them in large numbers. "Companies are using this as an assistant, a colleague," said IBM's Saft. "This will empower more people, like to write software," added Quennesson, who said AI thus could reduce the divide between technology haves and have-nots.
As for the Terminator/Skynet threat? Cheyer insisted that it's no more likely to happen than aliens attacking us. "We'll be hit by asteroids, more likely," he said reassuringly.
