Cloud computing titans battle for AI supremacy – Financial Times


December 5, 2016 Facebook Twitter LinkedIn Google+ Uncategorized


A new battle is breaking out among the titans of cloud computing, as competition in the fast-growing sector shifts to a fresh arena: artificial intelligence.

Amazon is the latest tech company to join the fray, boosting its focus on artificial intelligence at a time when Amazon Web Services, its cloud computing division, is locked in competition with Microsoft and Google.

AWS, which sells computing and data storage services to IT professionals, is the fastest growing and most profitable part of Amazon. Sales in the past 12 months grew 60 per cent to $11bn, making Amazon’s cloud business the largest in the world.

Amazon has embarked on a fresh push in artificial intelligence, recently unveiling three new AI services and investing in a deep learning framework, MXNet, which will power its future AI efforts.

Amazon already uses artificial intelligence in many aspects of its business — from ordering robots around warehouses, to identifying scenes in movies, to powering Alexa, its voice-recognition service. However researchers say that the company lags behind its rivals when it comes to pure research in artificial intelligence.

“They are playing catch up,” says Oren Etzioni, a computer scientist who heads the Allen Institute for Artificial Intelligence. Google and Microsoft have for years led the field in terms of AI research, he said, while Amazon and Apple have trailed behind them, but are now waking up to the importance of building their AI capabilities.

“In fairness, they are getting stronger,” he added, referring to Amazon. “But the question is how long it will take them to reach where Google and Facebook are, because Google and Facebook aren’t standing still either.”

Amazon says it has “thousands” of engineers working on artificial intelligence, but the company has been secretive about these programmes and has published few scientific papers about its AI efforts.

Last week, Amazon Web Services launched three new artificial intelligence services: one for image recognition, one for natural language understanding and one that helps a computer to read text out loud. While these services were already in use internally — for example Lex, which processes natural language, is used by Amazon’s Alexa — the launch marks the first time they will be available to outside developers.

“It’s not something that we talk about a lot, but we’ve found tremendous success internally with machine learning,” says Matt Wood, head of product strategy at Amazon Web Services. He says AWS is already working on more AI services, which will be launched next year.

However, one of the challenges has been that Amazon’s AI work is scattered in small teams throughout the company — for example, the engineers working on book recommendations may have little contact with those working on facial recognition software — making pure research difficult.

Amazon has considered building its AI capabilities through acquisitions, but has yet to make any big deals. Earlier this year the company examined Turi, an AI group based in Seattle, but Apple won the bidding process instead, paying around $200m for the start-up.

Mastering artificial intelligence is critical for Amazon because AI will be a key source of demand for cloud computing as more complex algorithms require more computational power.

“The workload for these data sciences services is going to be a big money generator,” says Al Hilwa, software analyst at IDC, referring to AI services that can be applied to large amounts of data. He estimates that less than 1 per cent of AWS’s workload today is based on these types of service, but expects this to rise quickly. “That is why everyone is falling over themselves to offer capability for data science,” he says.

Even though Amazon is still playing catch up in the science of AI, Mr Hilwa says the company has an advantage given the reach of AWS, which is much larger than the cloud divisions at Google or Microsoft in terms of revenues.

“You can expect the adoption to come faster because of the critical mass that they have in the cloud,” he says, referring to Amazon’s AI services.

Comments