I'll Be Back: The Return of Artificial Intelligence

Google, Facebook, Amazon spur rebirth of industry after decades of little corporate attention

Terminator Salvation, 2009.

Terminator Salvation, 2009.

Photographer: Richard Foreman Jr./Warner Bros./ Everett Collection

The artificial-intelligence industry, a field that conjures up images of humanoid robots and self-aware computer systems, is making a comeback at Silicon Valley companies like Scaled Inference Inc.

Inside a sparsely decorated office at the eight-month-old startup, founder Olcan Sercinoglu is developing an AI system that can help predict events, such as what website you’ll read next. While that’s far from the kind of AI found in science-fiction movies, there’s plenty of interest in this new generation of AI tools, which can help with everything from recognizing speech patterns to sorting through thousands of photos.

“There was interest to invest from almost all of the investors that I spoke to,” said Sercinoglu, who raised $13.6 million in funding for his venture in just five months last year after spending 13 years on-and-off at Google Inc. working on AI projects. “Many AI companies have raised significant money without any product plans.”

Scaled Inference is one of more than a dozen startups now forming the backbone of a mini-boom in AI, which is software that exhibits the intuition of people. After two decades of the field suffering from scant research funding and little corporate attention, a rebirth is being spurred by interest from Google, Facebook Inc., Amazon.com Inc. and others, with Alibaba Group Holding Ltd. Chairman Jack Ma saying this week that the Chinese e-commerce company will invest significantly in the area. In addition, falling technology costs have made the numerous computations underlying AI cheaper to perform.

In total, 16 AI companies were funded for the first time in 2014, up from two in 2010, according to data compiled by researcher CB Insights for Bloomberg News. The amount invested in the startups—some of which describe themselves as doing machine learning or deep learning—soared to $309.2 million last year, up more than 20-fold from $14.9 million in 2010. 

“The field was pretty static for 20 or 30 years,” said venture capitalist Vinod Khosla, who has invested in young AI startups including Scaled Inference and MetaMind Inc. Yet with advancements in computing power and the growing volume of digital information, “the promise has gone up,” he said.

Like Scaled Inference, many of the AI startups aren’t developing Skynet-like systems or human-like machines and instead are focused on making clever tools to solve specific corporate problems. Startup Expect Labs Inc. is creating software so retailers can add speech-processing capabilities to mobile applications. Clarifai Inc. has developed image-recognition systems that can sort through thousands of photos taken by wedding photographers to select the prettiest pictures.

Yet the startups face long odds, given a history of AI flops. In the 1970s, AI funding plunged after the U.S. and U.K. governments became fed up with the slow progress on efforts like autonomous robots. In the 1980s, the market for AI hardware collapsed further when personal computers from Apple Inc. and International Business Machines Corp. became popular.

“The thing I worry about most is over-hype,” Khosla said.

Behind much of the proliferation of AI startups are large companies such as Google, Microsoft Corp., and Amazon, which have quietly built up AI capabilities over the past decade to handle enormous sets of data and make predictions, like which ad someone is more likely to click on. Starting in the mid-2000s, the companies resurrected AI techniques developed in the 1980s, paired them with powerful computers and started making money.

Their efforts have resulted in products like Apple’s chirpy assistant Siri and Google’s self-driving cars. It has also spurred deal-making, with Facebook acquiring voice-recognition AI startup Wit.ai last month and Google buying DeepMind Technologies Ltd. in January 2014.

For Google, “the biggest thing will be artificial intelligence,” Chairman Eric Schmidt said last year in an interview with Bloomberg Television’s Emily Chang.

The companies have hired AI academics and engineers, some of whom are now spawning their own startups. Apart from Scaled Inference, former Googlers also founded AI company Moloco Inc. and Petametrics Inc.’s LiftIgniter. Amazon, Microsoft and Google engineers are also in key roles at startups including Sentient Technologies Holdings Ltd., SignalSense Inc. and Clarifai.

“At the big tech companies, you’re working with data at a scale that forces you to solve different problems than you would otherwise,” said Adam Berenzweig, chief technology officer at Clarifai in New York and a former Googler.

Falling technology costs have also helped trigger AI startups. Processing large amounts of data—a foundation of AI work—is cheaper as computer chips have gotten faster and the prices of storing and accessing data have dropped.

Systems called graphical processing units from companies including Nvidia Corp.—which enable companies to take on more of the monotonous-but-numerous computations needed to train and develop AI systems—have also dropped in price as GPUs have been manufactured on a mass scale for gadgets such as video-game consoles.

“The increase in performance over the last 17 years has been extraordinary,” said Jon Peddie of semiconductor market-research firm Jon Peddie Research. In 2014, people could buy a video card that was 84.3 times the performance of one from 2004 for the same price, he said.

The AI boom has also been stoked by universities, which have noticed the commercial success of AI at places like Google and taken advantage of falling hardware costs to do more research and collaborate with closely held companies.

Last November, the University of California at San Francisco began working with Palo Alto, California-based MetaMind on two projects: one to spot prostate cancer and the other to predict what may happen to a patient after reaching a hospital’s intensive care unit so that staff can more quickly tailor their approach to the person.

Theresa O’Brien, an associate chancellor at UCSF, said the university teamed up with the startup—the first such collaboration she’s aware of—because it wants to develop better approaches to bespoke medical treatment by employing computers to sort and link data, which AI can help.

MetaMind makes publicly available an interface to its deep learning architecture. On their website, one researcher trained an algorithm to distinguish among 4 species of malaria parasites, using just 62 sample images.
MetaMind makes publicly available an interface to its deep learning architecture. On their website, one researcher trained an algorithm to distinguish among 4 species of malaria parasites, using just 62 sample images.
Source: MetaMind

As more AI startups emerge, companies are turning into customers. American Express Co. said it has increasingly been using AI techniques to help automatically spot fraudulent transactions.

“Our machine learning models help protect $1 trillion in charge volume every year, making the decision in less than 2 milliseconds,” Vernon Marshall, American Express’s functional risk officer, wrote in an e-mail, without disclosing which AI companies it works with. “We have been delighted with how well this technology can detect fraud.”

Revenue—which has been in short supply for AI makers in the past—is now trickling in. Sentient, founded in 2007, is charging clients for technology that uses evolutionary algorithms to continuously ingest information and solve quantifiable problems, like how to profitably trade shares on the stock market. The San Francisco-based company closed $103.5 million in financing in November and anticipates having 100 employees by the end of the year, up from 60 workers now, said Sentient Chief Executive Officer Antoine Blondeau.

While Sentient originally made trading systems for its own purposes from its AI technology, it has since applied it to other industries, including using the system to evaluate a hospital’s data to predict the chance of a patient falling into sepsis.

“We expect to grow significantly,” said Blondeau, who declined to disclose Sentient’s revenue or how much it charges for its products. “This is a very nascent space where disruption can be very quick and because of the nature of the problems that are being addressed in a specific case, is highly lucrative.”