One Year Later: Now There Are Three

Way back in the middle of 2017, my colleague Jeff Fried participated in an exclusive member webinar with the Cognitive Computing Consortium on Leaders in Cognitive. He discussed the topic of, “The Race is On: Comparing Google and Microsoft’s Cognitive Services.”  I know… how is the middle of 2017 “way back”?  Well, in the terms of cognitive computing, AI, and machine learning, it seems that just last week can seem like “way back”.

The pace of change in this area is unprecedented, which prompted me to look back to see what was exciting then, what’s exciting now, and what some of the major changes are in this very promising slice of our market.

What was exciting then?

Jeff coined the term “cognitive arms race” in his presentation, and I really like that term.  It very much describes what is going on.  He focused on the offerings from Microsoft and Google specifically, covering agents, bots, cognitive capability, machine learning, and augmented reality.  I think it’s plain to see that both companies are taking a build and acquire strategy (Read more about the “Race for AI”, and the frenzied pace of acquisitions of AI startups here), where they are investing heavily in building capability, while also acquiring promising technology that helps them fill out the overall portfolio.  You can see evidence of this in what can come off as fragmented offerings, whose roadmaps come together down the road.  That’s a typical path for acquired capabilities.  Jeff had some great examples, and a clear consensus that the goal of both companies at that time was to win developers.  It’s clear that the more they can get their capabilities integrated into the work of others, then the better off they will be.  I think Jeff was spot on when he summarized that there is low risk in jumping in, that you should expect opaqueness in the implementations, and that you shouldn’t try to build this stuff yourself. Just wait a few months if something you need isn’t there yet.

What’s exciting now?

So, which shiny new capabilities have emerged in the vast amount of time of less than a year?  There has been a slew of announcements out of both Microsoft and Google, and I’ve highlighted a few that have specifically piqued my interest:


  • Public preview of Custom Vision service
    Microsoft is innovating fast in this area, with this announcement bringing state-of-the-art machine learning that offers developers the ability to train their own classifier to recognize what matters to them.  Add to that the continued increase in capabilities of the Face API, and it seems we’re one step closer to Spielberg’s vision of individualized facial recognition-based advertising, as he showed way-way back in 2002s “Minority Report”.
  • Microsoft’s Multi-headed search monster grows with Bing Entity Search
    Microsoft continues to have multiple irons in the fire where search is concerned.  These multiple strategies can generate some confusion, taking time to understand O365 search vs. AzureSearch vs. Bing for business vs. SharePoint search, etc. Even with these varying strategies, continued investment across them all is a good sign. Bing Entity Search allows applications to bring rich context about people, places, things and local businesses to drive engaging user experiences.


  • Open source release of DeepVariant
    The capabilities coming out of DeepVariant are exciting for those in the genomics space, but in general terms this speaks to the continued spread of AI and ML capabilities into specialized applications.  Seeing the general concepts deliver greater speed and accuracy into a highly specialized field like Genomics speaks to the flexibility of the frameworks that are being created.  I see similarity to the technical boom that the “space race” created for mankind and expect a similar boom from the “cognitive arms race”.
  • Google Acquires “AIMatter”
    In another play focused on mobile integration of vision-esque capabilities, Google folded in AIMatter, a neural network-based AI platform and SDK that rapidly detects and processes images on mobile devices.  This acquisition seems to indicate some interest from Google in next generation social apps, but time will tell as to where we see them use this technology.
  • More Features and More Power
    As expected, features and available computing power continue to reach the market.  Google’s TensorFlow 1.5 release brings Eager Execution and TensorFlow Lite as highlights, while the release of GPUs in Kubernetes into beta brings more raw computing power to ML workloads in the cloud.

Amazon will not be left out

Amazon should not be ignored in this space, given that they are as serious about winning the hearts and minds of developers as Google and Microsoft.  In fact, there are more “in production” machine learning applications deployed on AWS than any other platform.  I think it’s safe to say that Amazon is playing big and loudly in this area.  In terms of capabilities, the highlights below just scratch the surface.

  • Amazon SageMaker
    This is Amazon’s trainable framework that allows Data Scientists to deploy machine learning capabilities quickly on a common framework.  It follows a build-train-tune-deploy model, which cuts cycle time and speeds time to value.
  • AWS DeepLens
    The world’s first deep learning enabled video camera that is integrated with SageMaker and many other AWS services.  If someone hasn’t already created a HAL 9000 knock off with this, then I’m disappointed in the development community.
  • Amazon Lex
    This is Amazon’s entrant into the chat bot community, enabling both voice and text.  The concept here is to put the power of Amazon Alexa into the hands of developers.  If you’ve not interacted with Alexa, I highly recommend that you try it out.  My six-year-old talks to Alexa like she’s one of his friends.  If that’s not winning technology, then I don’t know what is.
  • Amazon Deep Learning AMIs
    The AWS Deep Learning AMIs aim to provide infrastructure and tools to accelerate deep learning in the cloud.  These are pre-installed with Google’s open-sourced TensorFlow and Microsoft’s Cognitive Toolkit (and a bunch of other frameworks) when deployed on AWS, so Amazon is betting on the “better together” concept in this area.  Not a bad approach when you can deliver your own capabilities and still be a part of making others’ capabilities better.  Maybe Amazon is the 3M of cognitive computing.

I’ll end with a bold prediction – the “cognitive arms race” ends in a tie!  Our bet is that Google, Microsoft, and Amazon will continue to innovate and acquire until they’re the only ones left.  In terms of the cognitive search market, the chance of any vendor in our space investing and competing with them is very low.  That is why our core strategy is to integrate quickly, integrate often and get this technology into the hands of users (Read more about that here). Engineering investment by firms in our market on proprietary AI or ML capabilities simply won’t be able to keep pace, causing companies who implement that closed technology to fall behind. We have the same belief about search engines.  With the investment and innovation coming from the likes of Elastic and Microsoft, a proprietary search engine simply won’t be able to keep up.  We believe in openness, flexibility, and best of breed AI/ML capabilities.