I recently watched a lecture by Dr Andrew Glassner on the subject of Deep Learning, and he highlighted something that I am not sure that businesses or governments really understand.
He pointed out that thanks to a number of highly intelligent and committed engineers and data scientists, the tools and techniques of artificial intelligence are there for the taking. The worlds most powerful tools are there for anyone to download and use as they see fit.
One might wonder then, why it is that every company and organisation does not have deep learning systems interrogating all of their data and producing knowledge and insights everywhere.
Glassner put his finger on part of the answer to that. He says that a significant problem is getting the data to the place where it can be used by these tools. This, he said, was actually a bigger problem than setting up the deep learning infrastructure to operate on the data. It is about getting the data to the people that need it in the format that they require.
In an average organisation a person searching for data confronts all kinds of obstacles, many of which are placed there intentionally.
The problem is that it is not wise to allow all of your data to be discoverable by everyone. There is so much that could go wrong in such a scenario. There are issues of licensing, copyright, privacy, attribution and so on that make the organisation vulnerable to law-suites or compensation claims. There is also the possibility of theft. What if your data finds their way into the hands of the competition. If your data are available for use are they not also available for misuse?
This is one giant complication in the management of data, but it is even bigger than it first appears. Data are like dangerous chemicals, they can leak into into other data, they can mix in and hide themselves away, in a map or on a chart with lots of other material, waiting for the day when they are recognised with horror by someone who expected that that stuff to stay confidential. This is often a problem that is too difficult to deal with. So the data with all of its potential, its treasures, stays hidden.
Written by Trevor Christie-Taylor