Microsoft announced incremental improvements to its best in class offerings for enterprises and knowledge workers in a session that demonstrated its lead in corporate AI services but further highlighted to me how slow the pace of AI development really is.
Microsoft’s offering is based on deploying AI trained in the cloud onto local devices to create richer services that can meaningfully improve productivity. It is in the context of productivity improvements that I have assessed its announcements and Build 2018.
1. Project Brainwave: a move by Microsoft to run its AIs at the edge not on graphics chips (GPUs) as everyone else does but on FPGAs – programable chips that can execute code much more efficiently than running the code purely in software, but are quite expensive.
However, given the huge requirement for AI processing at the moment, GPUs are in huge demand, meaning that FPGAs are currently a viable and economic alternative. This will be music to Xilinx and Altera (owned by Intel) who between them control the market for FPGAs but bad news for NVIDIA which dominates the GPU market. If this catches on, it will put pressure on Nvidia’s 60%+ gross margins as it will have to drop its prices to make FPGAs uneconomically viable again.
2. The resurrection of Kinect: not as a consumer device, but as a reference design with hardware and software that can be used by developers. Many of Microsoft’s demonstrations used sensors that either mapped a room or environment in 3D, interpreted video, still images, speech or the audio environment.
Old Kinect units from the Xbox have popped up in other use cases, making the creation of a reference hardware something that is likely to gain some traction. Creating a reference design is not about selling hardware units but instead is about making Azure the best place to deploy businesses processes. It is from this that Microsoft will make money.
3. Mixed reality: Augmented reality (AR, or ‘MR’ as Microsoft calls it) has been a massive disappointment in consumer, but there are plenty of applications in the enterprise. Now that Microsoft is almost entirely focused on the enterprise, HoloLens is a product that is ready to be used here and now. Gone are the days of Joe Belfiore shooting aliens from behind a couch – now it’s all about corporate use cases.
Unsurprisingly, the top use cases are remote assistance, special planning, training and development and collaboration. Remote assistance and special planning are by far the biggest of the four which is why Microsoft launched ready-made services for companies (remote assist and Microsoft layout).
Again, this is not about selling HoloLens units but making Azure and Microsoft 365 a compelling place to run one’s company which is where Microsoft will cash in.
4. Alexa and Cortana speak to each other: Alexa and Cortana are able to invoke one another which was demonstrated for the first time. Alexa in the kitchen was used to invoke Cortana to send emails while Cortana on the desktop was used to invoke Alexa to book an Uber and turn off a light. Very little has been developed beyond the original scope of this alliance and I still think that Amazon got by far the best of it. Very few people are likely to invoke Cortana from Alexa, but using Alexa on a desktop PC could get some traction. Furthermore, Amazon is stomping all over Microsoft’s territory with its moves into the enterprise which look to me to be in direct competition with Cortana.
In terms of improving the AI that sits behind these digital assistants, not much has changed confirming my belief that the pace improvement of AI remains glacial despite what the marketers would have us believe.
The net result is that Microsoft is doing exactly what it needs to cement its position as the provider of a place for companies to transact their digital work. To me, this confirms my rating of Microsoft as the global No. 4 in AI. It has moved beyond stage 1 (data capture and understanding) and is working on stage 2 where it creates services and applications from the data that it has processed.
This was everywhere in its keynote from smart site inspections flagging up anomalies to smart meetings which could identify, transcribe and translate participants both near and far as well as automatically create meeting action items.
Microsoft is far behind Google and Baidu in AI, but importantly it remains well ahead of Amazon and Alibaba in my opinion. These are its real competitors for the provision of enterprise services, and as a result I think that it can continue to grow faster than AWS and close the huge gap in terms of market share.
Microsoft has continued to have a great run, but even I am beginning to think that the valuation looks a little stretched. I would not be averse to taking some profits.
This article was originally published on RadioFreeMobile