How Microsoft is making the machine learning capabilities behind Cortana accessible to any app with Project Oxford
Cortana for all: Microsoft's plan to put voice recognition behind anything — Microsoft and co. make computer vision, voice, and text processing a Web request away. Tweets: @timoreilly Tweets: Tim O'Reilly / @timoreilly : Microsoft is turning Cortana into a platform http://arstechnica.com/... #ProjectOxford beating Google & Apple to a huge opportunity
Context & Ripple Effects
In May 2015 Microsoft did something Google and Apple had not yet done with their own assistants: it took the vision, voice, and text processing that powers Cortana's machine learning and exposed it as web APIs any developer could call — Tim O'Reilly flagged it at launch as Microsoft 'turning Cortana into a platform' and 'beating Google & Apple to a huge opportunity.' The move reframes an assistant feature as infrastructure.
The subsequent arc shows the platform bet compounding even as the consumer assistant struggled: new API previews for speaker recognition and video arrived by December 2015, the Cortana Skills Kit opened voice-app development by 2017, and by CES 2018 Microsoft was explicitly playing 'the long game' after Amazon and Google overshadowed Cortana on smart home and SDK news.
First-order effects
- Any app developer gains immediate access to computer vision, speech, and language models through a web request, removing the need to build or host ML capability themselves.
Second-order effects
- Google and Apple face pressure to open their own assistant technology on comparable terms, since Microsoft has made its ML stack a distribution channel rather than just a product differentiator.
- Microsoft's own roadmap follows the demand: the API catalog expands within months into speaker recognition and video, turning Project Oxford from a demo into a growing service line.
Third-order effects
- Assistants shift from end-user products toward developer platforms — the structure behind the 2017 Skills Kit and chatbot strategy Microsoft staff later described — where value accrues to whoever hosts the intelligence layer.
- The pattern anticipates today's AI infrastructure platformization: model capability sold as metered cloud services, with consumer-facing assistants like Cortana ultimately mattering less than the underlying APIs.
The trend: AI capability is migrating from embedded product features to rentable cloud platforms, with Microsoft an early mover in turning assistant internals into developer infrastructure.