Microsoft Cortana vs IBM Watson

2 thoughts on “Microsoft Cortana vs IBM Watson”

  1. First of all thanks for the insight! I wanted to see if you could elaborate more on the definition of “scriipted’ when you are referring Cortana. Also, is there any documentation out there that dives deep into the technological differences between the two? Somehting that explains how they can compliment each other? Maybe what Watson can do, but Cortana can’t and vice versa. Thanks!

    1. Hi Justin, first a quick sorry for the late reply.

      It is Deborah that states that it is scripted and I interpret it as her content team is actually writing all the dialogues for Cortana, but do look at the above video to find out what exactly she said. To be fair to Microsoft in this comparison they have other cognitive solutions that are more “cognitive”, which she probably should have mentioned in her talk. Naturally there are benefits and challenges with both and I think it is a matter of company preferences which one you chose.

      I have not found a high quality analysis of the differences between Watson and MS Cognitive Services unfortunately, you can find fragmented details out there. If you have Gartner access, I assume they have written a report on it of similar.

      Most seem to state that IBM have a head start given their focus on this for so many years. MS, Google, Amazon and to some extent Facebook is in the mix as well, but for the large enterprise I would say IBM is still in the lead.

      Similarities in several areas, but a few differences I can list are the following:

      1. Retrieve and Rank
      This (in large parts) is the part of Watson that won Jeopardy. It is the API that learns & reason in a given domain (after initial teaching by domain experts). The uniqueness is definitely in the ranking part I would say. This is “the brain” and it’s capabilities to ingest, understand, learn, reason and even suggest actions / decisions (incl evidence) is impressive to say the least.

      2. Image recognition
      Watson is the only solution that has the capability to train your own classifier. This might sound as a small features, but when you think of it, it is essential for most companies to have this functionality.

      IBM also just passed Microsoft in speech recognition, with an accuracy of 5.5% fault rate. MS was at 5.9% prev. Humans are by most stated to be at 5.9% IBM claims human errors are at only 5.1%.

      Those two are from the top of my head. Given my limited knowledge of MS it is hard for me to go into details on what are the biggest benefits of their Cognitive APIs. Sorry for that.

      As you see above I have not mentioned bots or other interfaces, but focused on the cognitive parts. Bots are just another interface to the information. Some come can understand natural language and classify, some is purely scripted etc, but the power of cognitive is not in the interface, so I often try to leave that out.

      If you have further thoughts I am happy to continue the discussion.

Leave a Reply