Navigating Names at Numenta
For a company that is launching its first product, we have an unusual number of terms and names, including Numenta, Grok, HTM, and NuPIC. This abundance of names reflects our unusual mission. I’m writing this blog post to explain how all these names fit together.
Our company was founded over nine years ago as Numenta, and we are returning to using this as our company name. Numenta’s mission is to lead the new era of machine intelligence. This mission is ambitious, but it’s one that we are proud of and deeply committed to, and we have made real progress towards its achievement. The origin of the word is from the Latin “mentis”, meaning “mind”, so a “new mind”.
In order to understand and make practical our machine intelligence technology, we now are building products on top of it. Consequently, Numenta is both a technology and a product company. We are using the name Grok for our first commercial product. We just launched Grok for IT Analytics on AWS (Amazon Web Services). We anticipate creating a variety of Grok products over time, all of which will use our core machine intelligence technology, but include additional functionality targeted at specific customer problems or domains. Our goal with Grok and other products is to become the leading platform for machine intelligence.
Once you look deeper into the technology behind Grok and our other applications, you will want to understand where the acronyms fit in. The broad theory described by Jeff Hawkins in his book, On Intelligence, is called Hierarchical Temporal Memory (HTM). HTM is a theory about key characteristics of neocortex, describing learning as a memory-prediction system operating on spatio-temporal data. HTM forms the theoretical underpinnings for our work. The learning algorithms within HTM learn sequences and find patterns in temporal data. These HTM learning algorithms are modeled very closely after biological principles, and can be mapped to a single layer of neurons in the cortex.
In 2013, we created an open source project implementing HTM in order to further our mission of being a leader in the new era of machine intelligence. We refer to the open source project as NuPIC (Numenta Platform for Intelligent Computing), and it is hosted at http://numenta.org.
So, that’s all there is to it! In brief:
|Numenta, Inc.||Company name, including both technology and products|
|Grok||Name for first commercial product, Grok for IT Analytics|
|Hierarchical Temporal Memory (HTM)||Theory of neocortex described by Jeff Hawkins in On Intelligence. Includes learning algorithms that find patterns in streaming data, predict next steps, and find anomalies; a core building block for HTM.|
|NuPIC (Numenta Platform for Intelligent Computing)||Open source project where HTM is being explored and advanced by the developer community|
|http://numenta.org||The home for NuPIC, the open source project|
|http://numenta.com||The home for Numenta, the company, and the Grok product|
As I said up front, Numenta is an unusual company. We work with scientists who are mainly interested in the HTM theory. We work with developers who are mainly interested in our open source project using HTM. And we work with customers who are mainly interested in the Grok product built on top of these technologies, and prefer to think of the underlying technology as a black box. By using terminology to differentiate these audiences, it makes life a bit more complex for us, but we believe that it makes our work and our direction far more clear.