Artificial intelligence is advancing. Medical diagnosis, stock trading, even cow milking are applications being applied to artificial intelligence techniques today. How long before a robot takes over your job?
The answer to that question depends on what your job actually entails. Last month I was having dinner at this rather attractive branch of the chain restaurant Chili’s. Taking a seat, I noticed a screen on the table. The “Ziosk”, nifty device, let me order my dinner, pick from a selection of drinks and pay for it, all without any human interaction. I could even summon another drink with a push of a button. The waiter, whose role was reduced to ferrying dishes from the kitchen to my table cracked a joke about one day soon his job being completely replaced by a robot. I think he may be right.
If your job is data operations, how soon might it be before you become as concerned as the Chili’s waiter? Well, one thing in common with regular systems and artificially intelligent systems is they both act on data inputs. A routine transaction processing system that has been around for decades and an artificially intelligent system of the future can both go badly wrong if fed with bad data. Every day, all over the industry, a huge amount of manual effort goes into tedious work to make data ready to be used correctly by such systems. This is true for many applications in the industry, particularly scrubbing of symbol changes and corporate actions. A huge amount of effort is also expended in re-applying edits that have already been decided upon. If a decision is made to change a piece of data, rather than just re-do that change again tomorrow, it is much more efficient if the data management system decides whether to apply it again or if something relevant has changed and then make a decision itself about whether to apply that change or call on a human for advice.
These type of applications have an interesting side-effect – the extra data that the decisions themselves generate. If a data management system is routinely changing a rating on a bond from BBB to AA and then it decides that something relevant has happened that makes that automatic change unreliable or questionable, that information in itself is valuable. Somebody else could find that information useful in another decision. One day, data vendors may publish consolidated feeds of decisions that have been taken across the industry about their reference data rather than just plain old feeds of actual reference data values. Until then, I think I’ll have another beer with my chicken fried steak — click.