LBB's Laura Swinton on how Microsoft stacks up against Google, Amazon and Facebook as we approach the intelligence age
“There are two companies creating super computers: Google and Microsoft, because we’re indexing the entire world,” says Steve Sirich, GM of Bing Ads at Microsoft. “We’re creating a digital canvas of entities – people, places and things – and in our intelligent corpus of information, which is obviously powered by Bing, we’ve got billions of entities of information.”
By 2020, it’s predicted that humanity will have generated 44 zettabytes of digital data – that’s 44 trillion gigabytes. That’s every digital photo, every online search enquiry, every Facebook post, every academic article, every LBB post. And the future belongs not to those who can simply sort and categorise that data, but whoever can actively connect and find meaning in that data. Microsoft reckons it could be them.
“When you think about the IoT and our ability to process interaction across so many different devices and services it’s going to require artificial intelligence to do that, because we don’t have the capacity to do that as humans,” explains David, who feels like we’re approaching a ‘tipping point’ where AI will be a social necessity.
And Microsoft has slowly, quietly, smartly been preparing itself for just this point. For years the nerdier, less sexy computing stalwart to Google’s zeitgeisty cool kid, Microsoft’s various elements are coming together and a cohesive, powerful vision of their future is emerging. According to Wired writer Jessi Hempel, Microsoft is experiencing an ‘AI comeback’ and can more than hold its own against the big tech players angling for a slice of the future – and that’s thanks, in part, to Satya Nadella’s promotion to CEO. Whether it’s their Azure cloud computing service, Windows, Office, the search power of Bing, the communications tools of Skype, or, indeed, the intelligence of Cortana, AI cuts across everything.
This month, Microsoft even announced that it was taking on Google’s AI capabilities by establishing its own specialised AI research lab. To date, Google’s Deep Mind has been the darling of the armchair artificial intelligence expert, ever since its victory over high ranking Go player Lee Se-dol in March 2016, and it’s been working towards a ‘general AI’ with Google Brain. Microsoft will bring together over 100 scientists, including cognitive psychologists, to try to create a general AI – one that is flexible and can learn how to do work in different areas rather than focusing on one specialised function.
Another part of Microsoft’s strategy for AI supremacy is democratisation. They’re also empowering outside developers to build bots and apps that can put all this information and AI firepower to use – at the moment there are about 130,000 external developers using these tools. And they’re not just building bots for the Microsoft ecosystem, but for the likes Android, iOS and Facebook too. These developer tools are offered up as cognitive services (which mirror individual functions of the human brain) – so there’s vision (e.g. facial recognition, emotion identification, a content moderator that can identify NSFW images), language, speech, knowledge and search. This idea of creating souped up analogues to human cognition is captured rather nicely in the Seeing AI app. This was created by a Microsoft developer called Saqib Shaikh, who is visually impaired. In 2016 he revealed a prototype app for smartphones and Pivothead Smartglasses that was able to recognise and describe action playing out around the user and read text aloud. Seeing AI was released in 2017 for iOS.
And what of Bing, the not-so-sexy search engine? It turns out to be the organisational glue that ties the whole thing together. Last September, Microsoft announced that it was combining its research division, Cortana division and Bing division into a 5,000 strong behemoth of an AI division. And Bing’s inclusion here isn’t just about AI-powered search and the opportunities for image or voice led search. Bing can also serve as a useful tool for collecting behavioural data to feed the beast; as data scientist Seth Stephens-Davidowitz reveals in his book ‘Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are’, our online searches can be pretty revealing. As an example, a study by Microsoft researchers published last year in the Journal of Oncology Practice looked at the collective search history of a large number of pancreatic cancer patients and found patterns in the searches that they think could help identify between 5 and 15% of sufferers in early stages of the disease. It’s a project that reveals a real glimpse of the power that the vast data sets from a search engine like Bing could have when combined with AI.
For those that subscribe to the Elon Musk and Stephen Hawking school of thought - that AI spells doom for humanity - Microsoft has been careful to lay out strict parameters for its artificial intelligence projects. Like a latter day Isaac Asimov (the author who penned the famous Three Laws of Robotics), last year Satya Nadella set out six design principles of AI. These include the following: AI must be designed to assist humanity; it must be transparent; it must not destroy the dignity of people; it must be designed for ‘intelligent privacy’; it must have ‘algorithmic accountability’ which would allow people to undo any unintended harm; and it must guard against bias.
“We recognised that what we’re dealing with, I would suggest, is a corporate responsibility that we have with AI. Beginning with Satya, he definitely wants to foster a culture where we’re innovating and experimenting fast but we’re doing so in a very responsible way,” says Steve. “Its very much about the culture of who we are. Our culture and our mission are very much about empowerment of individuals and organisations to achieve more. But the way to do that is to build trust and trust is built when you’re representing the truth and you’re representing in inclusiveness and you’re representing all classes and you’re not biasing.”
This all being said, despite the computing heft, smart data and technology, Microsoft’s journey towards AI primacy has not been without hiccups. Their chatbot Tay, originally devised as a teenage girl persona who would learn from social media interactions, soon morphed into a racist, misogynistic troll once she came into contact with the Twittersphere. But to their credit, the failed experiment still became a learning opportunity as it seems to have informed Satya’s principles. The broader challenge, when it comes to AI and diversity and inclusion, is that AIs are driven by the biases (unconscious or otherwise) of the humans who design them and the information they are fed. That’s why Satya’s design principles are so important.
Looking forward, the next step in Microsoft’s AI push will be the launch of Invoke in the Autumn, a speaker that will take Cortana into your living room – and take Microsoft head-to-head with Amazon Echo and their AI assistant Alexa. It’s a piece of the puzzle that combines voice with Bing’s AI search capabilities and gives Cortana a platform beyond Windows PCs and laptops. However, given Amazon’s early bird infiltration, it’s predicted to corner 70.6% of the voice-enabled speaker market this year. With Google’s Home and Apple’s Siri Speaker also entering the market, Microsoft has a big fight ahead of it to ensure the kind of penetration it needs. On the other hand, Microsoft’s aforementioned democratised attitude might give it an edge. The Invoke is made by third party manufacturer Harman Kardon, and HP and Intel have indicated that they would be building devices for Cortana. Just as a plethora of manufacturers build computers that run Windows, the company hopes that a similar strategy will work for its voice AI.
There’s a lot of interesting stuff going on at Microsoft and as the sun rises on the Age of AI, it’s certainly gearing up to contend against Google, Amazon and Facebook.