Event Report - July 2017 Microsoft London AI Event

 

So, look at my musings on the event here: (if the video doesn’t show up, check here)

 

No time to watch – here is the 1-2 slide condensation (if the slide doesn’t show up, check here):
    Want to read on? 

Here you go: Always tough to pick the takeaways – but here are my Top 3: Microsoft gains momentum with broad AI portfolio – One can argue that Microsoft may have come late to the recent Machine Learning / AI market trend, but the vendor is certainly into the market now, with a broad, possibly one of the broadest offerings. AI changes almost all of Microsoft’s products, and the vendor has a broad product portfolio – so this does ultimately not come as a surprise at all. So, we see AI offerings coming to Microsoft Office, to Outlook all the way to the virtual keyboard offering SwiftKey (that Microsoft acquired February). Languages remain a border for communication and the online translation capability that Microsoft has built – are very good, embedded in to PowerPoint at first. Yours truly was able to test it with English, German and Italian – and it worked better than I expected. Especially as the translation model still operates on a generic universal (but likely monster size) pattern recognition model. It’s fascinating to see how well an older technology approach can do when pushed very hard (and on the flipside, shows that e.g. Apple – using the same architecture for Siri – has not done much to help her understand users). Speaking of Apple – Microsoft continues its ‘embrace & extend’ strategy for iOS and Android, also for AI, with a release of the “Seeing AI” product – a fun implementation of picture / pattern / emotion recognition to understand and see how the machines can make sense of the world around us. Finally, Microsoft goes all the way to the roots of its AI efforts, to Bing, where almost all of AI started and showed the Bing Entity Search API. This API, as well as the gesture API of Project Prague are interesting and important developer offerings to attract them to build next gen Applications on Microsoft cloud technology.

Beyond the direct product impact, Microsoft announced “AI for Earth” bringing the heft of R&D and deep pockets of Microsoft to the challenging environmental questions that the world sees – and where AI certainly can help. 
 

Packed Event

Dedicated AI team in Microsoft Research – Microsoft also announced the formation of a dedicated team within its research organization that will focus on a more integrative approach to AI. The team of more than 100 researchers has representation in all the company’s labs around the world, including  the UK, the location of the event. Good to see Microsoft adhering to good research principles when it comes to basic research (as it needs to happen with AI) – starting with a cross-disciplinary approach. Microsoft has the deep pockets to do this, one of the few vendors who can, and it will be interesting to see what the approach will yield. Given that there is a global arms race towards usage of AI for all kinds of computing questions, it will be key that over the multi-disciplinary approach Microsoft doesn’t not neglect speed to market. But that’s a common challenge Microsoft has been able to master well during its overall existence.
 
 

Chris Bishop announced AMLab partnership

Democratization of AI and Ethics – Always good to see the commitment to bring AI to more people and use cases, one of the leitmotivs of any Nadella presentation these days. The event was certainly in line with this and presented many examples for it – all starting with AI for Earth. And good to see that Microsoft has created an ethical design guide for all its employees working on AI. An important subject and mechanism amongst the ‘grown up’ AI vendors.
 

Great London Views

 

MyPOV

A notable event to attend, showing Microsoft’s across the board commitment and push to AI. AI everywhere is pretty much at play at Microsoft and is showing first and promising results. Having the event in London showed the R&D investment that Microsoft is doing beyond Redmond / the US and the UK certainly has a substantial contribution to AI in general and a significant Microsoft presence. But it is early days, e.g. one of the mechanisms that have always worked for Microsoft - connecting it all into a suite of consistent and user centric offerings has not happened yet. E.g. translating a slide with a picture of a person in PowerPoint, mentioning the name of the person - will not (yet) make it a recognized person in the “Seeing AI” app. And there are of course the downsides from a data privacy perspective – but Microsoft must realize that its competitors in the space are ruthless at leveraging those synergies. It is likely Microsoft needs (and will) do the same.

On the concern side, Microsoft has to find a match to the uptake in mindshare and real-world AI deployments that Google’s Tensorflow is seeing, if not Tensorflow will be the Kubernetes of 2017 – a Google launched open source offering that dominates an important aspect of the cloud infrastructure. And one can argue that Tensorflow is even bigger stakes than Kubernetes, as it not only dominates the AI / Machine Learning equation, but due to data gravity pulls along substantial data storage and processing load. Not to mention the model training and execution loads.

But for now, it is good to see the progress at Microsoft, who is pushing AI wide and deep across its existing products, as well as new offerings. A deep R&D and research check book certainly helps, but must show tangible, shorter horizon returns to capture what matters most in AI these days – cloud load – both from a storage and compute perspective. Stay tuned.


Want to learn more? Checkout the Storify collection below (if it doesn’t show up – check here).