Computer systems today are often called on to make predictions based on data provided to them. Sales forecasts, power usage, the weather, and even the little lights that come on in posher cars that tell you something is starting to go wonky are all based on software that is in turn based on predictive algorithms. This is how Amazon constantly changes what it thinks you might buy, and Goggle can creep you out by offering to sell you what you are searching about. The crafting of such programs are calls for laborious and highly skilled coding and constant fine tuning of the factors used to make the predictions.  Each system has to be customized for each situation and is very expensive to maintain. We are a long way away from the Star Trek world of telling the computer what you want analysed or having Mr. Spock work it out.

Perhaps we are not that far away after all, OGs (Original Geeks) know that Jeff Hawkins was the mastermind behind the first successful  handheld computer, the Palm Pilot, the grandfather of all Palm PDAs. He then went on to co develop the Treo, the first smartphone. Since 2005, he has been working with Numenta following what was actually his first love, neuroscience. Hawkins has since been seeking a solution that has dogged artificial intelligence developers since day one, how to make computers learn and understand without all the tweaking it takes now.  Grok, a prediction engine, now in private beta seeks to do just that.

Calling the cloud based system “Grok” is both a homage to the late Grand Master of science fiction, Robert A. Heinlein, who coined the phrase as a Martian word in his 1961 classic Stranger in a Strange Land   and as a thumbnail explaination of what Hawkins seeks to do. The Heinlein defines the word thus; “Grok means to understand so thoroughly that the observer becomes a part of the observed—to merge, blend, intermarry, lose identity in group experience.”

Grok seeks to meet its ambitious name by use of Cortical Learning Algorithms that allow the system to create models on the fly and modify them based on new data in a process called “online learning” Grok can also flag anomalous and unusual data as well, based on what it is fed. It is also far more generalist in its approach, being able to tailor itself to a specific situation. For example, if you are an engineer tasked to create building climate control systems, instead of custom programming a system for every building,  one instance of Grok can be developed for a building and then deployed to all similar buildings and Grok will learn the quirks of the particular building system its managing.

Its still early days yet, but Grok promises to lead to systems that make Siri developmentally delayed by comparison. And understand you all the time.