The 'summer of AI' is here, this startup chief says

24.06.2016
Artificial intelligence is still surrounded by an aura of mystery, and it would be tough to find a better illustration than the story in the news last week about a British grandmother who includes "please" and "thank you" in all her Google searches.

"Please translate these roman numerals mcmxcviii thank you," read the search request from May Ashworth that ultimately went viral when her grandson tweeted it on Twitter.

"I thought, well somebody's put it in, so you're thanking them," Ashworth reportedly explained. "I don't know how it works, to be honest. It's all a mystery to me."

It's an endearing tale that drew notice from Google itself, but it also underscores the way AI, in particular, has been something of a black box.

"Most people don't know how it works, so there's this sense of wonder where you ask, 'what could it do next'" said D. Scott Phoenix, co-founder of AI startup Vicarious.

That potential will soon unfold, Phoenix believes.

After a long winter, the "summer of AI" is now beginning, he said. "There's a couple of key mile markers we can look for, and some have already happened."

Already here is the availability of "lots and lots of data" along with a reduction in the price of storage, memory, and computing power to "near zero," Phoenix explained.

At the same time, it's now recognized that even tiny improvements achieved through AI -- an additional tenth of a percent in advertising click-through rates, for example -- can have a massive effect on revenue.

"Now there's this engine running, and we know AI matters," he said. "There's now a stable market demand."

Coming next is a new focus on the skills needed to improve the technology.

"We saw something similar in the early days of software -- and the early days of the Internet," Phoenix said. "We're entering that era of rapid improvement."

Vicarious is enshrouded in a fair bit of mystery itself. Founded in 2010, the company hopes to take AI beyond the specific-purpose systems seen today -- one piece of Google software focused on playing the game Go, for example -- to build a more general-purpose one.

But it's holding its cards close, and Phoenix says it's not yet ready to launch a product. 

"We're really interested in having a single, unified algorithmic architecture that works in concert to reach human-level performance across multiple domains," including vision, language, and motor control, Phoenix said.

"We're especially interested in embodiment," he added. "There's a lot of things you learn by being in a body and interacting with the world."

Toward those ends, Vicarious is drawing from neuroscience, deep architectures, and generative probabilistic models to create AI that requires "orders of magnitude" less training than traditional machine-learning techniques do, it says.

In 2013, Vicarious announced that its technology had gained the ability to reliably solve modern CAPTCHAs, thereby leaping the hurdle that's used as a sort of Turing test on many websites.

The company has raised roughly US$70 million to date, with investors including Facebook CEO Mark Zuckerberg, Tesla Motors CEO Elon Musk, Amazon CEO Jeff Bezos, Salesforce CEO Marc Benioff, and Box CEO Aaron Levie.

Phoenix's company may be working to create AI with a broader style of human-like intelligence that can learn and generalize, but he isn't worried about the widespread loss of jobs that's so commonly foretold.

"If you look back through past technology revolutions -- automated factories, software, PCs -- many jobs were destroyed by those, yet we still have jobs today," he said. "The job I have today didn't exist back then; others were very different."

The flip side of the story is that AI will help us do more than we did before, so jobs will change -- but not disappear.

"The idea that we'll run out of jobs or things for humans to do is, at least in the short term, pretty far-fetched," Phoenix said. "There are going to be lots of new types of jobs created."

One need only look around to see that "there's so much to be done, so much to be invented to get us to a better society," he added. "There's so much we can achieve by having our systems be more intelligent -- there's always more work we can do and higher heights we can reach as a species."

Katherine Noyes

Zur Startseite