Life on Mars…or what needs to be done to get AI right

23, February, 2017


Artificial Intelligence -


The past week has been one of particular announcements. Elon Musk unveiled his hotly anticipated plan to send humans to live – and die – on Mars, outlining plans for up to a million colonists living in a self-sustaining society, while Microsoft CEO Satya Nadella announced plans to add intelligence to core products like Office 365, Skype, and Dynamic. Unfortunately, both announcements left more questions than answers.

Life on Mars

Musk kept us in the dark about how his expensive missions would be funded and the technologies needed to extract human-sustaining resources from Mars. Similarly, Nadella ignored the fact that 90% of the world’s data is unstructured (IDC) and will not be indexed by Microsoft’s new machine learning solution. I don’t know about you, but I’d prefer my AI to take all its surroundings into account rather than having a walled garden view of the world.

While the point of Musk’s speech was to show us that there’s a reason and roadmap to interplanetary life – really big thinking, Nadella rehashed the same the same song about more productive tools and platforms – but only based on the data that Microsoft can gather from their own apps.

Heterogeneous world

We live in a heterogeneous world, structured and unstructured data is everywhere. The volume of data has increased and now lives across an ever-expanding array of enterprise locations. Nadella even jokes that we will soon create so much data that “we are getting to a point where we don’t even know what to name things.” But, amazingly, there is nothing about heterogeneity in Nadella’s plan. How can Cortana provide useful intelligence when over 90% of enterprise data is simply ignored?

When the web rose to prominence, Yahoo built a successful business by categorising the web and tagging sites based on specific categories. Unfortunately, this manual approach to tagging could not scale as the Internet exploded and Yahoo’s index was left void of a huge portion of the web. Enter Google and Larry Page who developed the PageRank system eliminating the need for human categorization that Yahoo was wedded to.

Microsoft’s homogenous AI bet could face the same destruction at scale risk as Yahoo. The winner in enterprise AI will be the provider of the first bridge between all unstructured and structured data.

Light years away

In a heterogeneous world, until there’s a bridge between all structured and unstructured data, machine learning just won’t have the holistic view of the world to deliver truly ground breaking AI. Without a bridge, AI will be limited to cute handwriting recognition and text prediction apps.

Who will it be? Watch Google’s Diane Greene and Salesforce’s Richard Socher’s moves in this space. As for Microsoft, this week’s announcement shows they are missing the point about machine learning.