Deep learning, technology that can mimic the human brain, has long been a cornerstone of research in the artificial intelligence community. It is considered so pioneering and important that Google paid $400 million last year to acquire a leader in the field called Deep Mind. So it was surprising that when a high-level panel of artificial intelligence experts gathered at a forum in Menlo Park, California one evening last week, the consensus opinion was that deep learning represents the most overhyped part of their profession.
The forum, sponsored by the Churchill Club and hosted by SRI International, featured highly-respected experts such as Adam Cheyer (co-inventor of Siri), Nigel Duffy (CTO of Sentient and a leading expert in machine learning) and Lauri Saft who is leading IBM’s Watson Partner program. They were joined by Kevin Quennesson (currently manager of Twitter Cortex and the developer of Apple’s first 3D iPhone game) and Norman Winarsky who is responsible for creating some of SRI’s highest-value ventures (including Siri).
Duffy pointed to recent press that hinted his company and others were able to duplicate the human mind as an example of how misguided the common understanding has become. “Fear derives from the notion that we’re producing this model of the brain,” said Duffy.
But the overhyped nature of artificial intelligence research into deep learning may be due, in part, to the secrecy surrounding new developments in the field. There has been growing speculation that the next major breakthroughs will come in the use of technology as personal assistants, which gained more credibility last week with the announcement from Facebook that they were testing a personal digital assistant inside of their Messenger platform.
This puts the social media company in an elite club. Apple, Google, and Microsoft are all reportedly working on personal digital assistant technologies as well.
There was also a significant funding development in the field when deep learning startup HyperVerge announced earlier this month that they had raised investment capital from VC firms New Enterprise Associates (NEA) and Milliways Ventures. NEA, one of the largest VC funds in the world, previously backed Salesforce and Juniper Networks. Milliways was an early investor in Facebook.
Whether it includes deep learning or not, the true game-changer in this field may be a stealth startup called Viv Labs. Founded by Cheyer (Siri’s creator), the company is reportedly working on a new program using voice recognition that will essentially be able to teach itself, rather than having to depend on engineers to program it.
Cheyer was careful during last week’s SRI forum not to reveal many details about Viv (which is named for the Latin root meaning “live”), but his comments were helpful in showing where the company may be headed.
“A lot of what we build is held together by handwritten software, and that hasn’t changed much,” said Cheyer. Siri’s creator went on to explain how artificial intelligence can play a “huge role” in changing how software is written, which could indicate that this will form much of the core technology surrounding Viv.
Cheyer also dismissed notions that artificial intelligence research will result in the rise of a “computer god” who will control our lives, whether we want it or not. “It is a true question to think about,” said Cheyer. “But it’s not something I’m afraid of. We’re more likely to be hit by asteroids.”
SRI’s Winarsky pointed out that artificial intelligence is being driven by the huge amount of data now being collected by connected devices in areas such as robotics, cars, and photo recognition. “Artificial intelligence is here and it’s a time for harvesting,” said Winarsky. The question now surrounding this rapidly moving field is how much of the intelligence gathering and learning will be done by humans or the computers they program.