Here’s a speculative question: What if you could quantify thought leadership?
It turns out you can — at least in a sense. To the extent that thought leadership leads to intellectual property, the existence and legal protection of that intellectual property tells an interesting story on a company-by-company basis.
Patents offer a relevant set of metrics to apply in considering this area. Companies with an unusually large number of patents have a strong, practical case to make as thought leaders. This is because if their ideas weren’t innovative, they wouldn’t hold patents in the first place; if their innovations weren’t frequent, the number of patents would be small.
Here at SparkCognition, we’re fortunate to work at a company that is committed to innovation — in fact, it’s one of our core values. You won’t be surprised to read that our innovation has led to a large and growing number of patents and patent filings.
At the moment, the company has more than a hundred patent filings, of which close to forty have been granted patents in the United States, more than fifty are pending, and another fifteen are pending internationally.
(Don’t take these numbers as etched in stone, though; they’re all moving up. Take them, instead, as etched in butter.)
Another relevant metric might be the time required before the United States Patent and Trademark Office responds to a filing by issuing an actual patent.
For all software-centric filing in the United States, the median time is 858 days before a patent is issued (if it is at all). For SparkCognition, it’s almost a full year faster: a median time of only 537 days. While the exact decision-making processes utilized by the patent office are not published to the outside world, it seems fair to say that the more clearly innovative a filing is, the faster it’s likely to result in a patent. That speaks well of our relatively high speed of patent issuance.
We might also consider the frequency of our patent filings. As SparkCognition has become more mature, has that frequency declined? No; it has increased. In 2020, the annual number of filings set a company record — 35, up by more than half compared to the previous record of 22. In 2021, expectations are that the record will again be broken via 37 or more filings.
These patents aren’t merely intellectual trophies either–they’re a reflection of our commitment to deep science and applied AI research that has been at the heart and soul of our company since its founding. The innovations they protect have been leveraged to create real value in the real world in many different contexts and categories, and the breadth of our research is a feature and not a bug. As any modern day practitioner of AI knows, what is initially developed for one task or domain might later be applied to others in novel ways that continually re-define the state-of-the-art.
Let’s focus on anomaly detection, for instance. In this category alone, SparkCognition holds nearly a dozen patents.
One such patent concerns expanding the reach of monitoring and maintenance capabilities. Broadly speaking, solutions that tackle this complex area assume modern, high-dollar assets replete with built-in sensors that continually track and report operating status in different areas like power consumption, performance, and others. Our patent focuses on the more everyday case of assets like pumps that lack such comprehensive sensorization. Via the method described in the patent, it becomes possible to generate a model of normal behavior for such assets based on historical data that is known not to contain any anomalies. Thus ongoing monitoring becomes possible for lower-ticket assets without going through a relatively costly process of training and sensor calibration.
Another instance improves the way organizations react when monitored devices generate an alert (due to circumstances that seem to require a response). The commonplace goal in such a scenario is to automate such processes wherever possible, driving a faster response, by searching raw feature data to determine the root cause of the problem. So far, so good…but many times such an approach won’t arrive at the correct assessment because the device can change over time due to issues like ongoing maintenance (or lack of it) or simply routine wear. This means the historical data, created earlier, may not be relevant, and the root cause of the current problem may not be the same now as it was for the older problem. Our patent delivers a more accurate assessment by considering featural importance metrics too — providing a broader understanding of the asset’s status and a quicker diagnosis and resolution of the true root cause.
A third innovative patent in our ever-expanding portfolio concerns the training data used to create and improve an AI model. At present, in many cases, obtaining such data is one of the slower and more expensive stages in getting such a model up and running and creating value for a client. What our patent provides is a more efficient way to address this stage. To do this, it draws label data from similar assets as those to be covered in the model. That data is subsequently augmented with engineered features and classified in a manner that takes into account the context of both asset groups, so as to ensure it applies correctly in the new scenario. Thus an accurate model can be built, yet the total costs and time required for training it are both substantially lower than via conventional methods.
All three patents pertain to our industry-leading SparkPredict® product, and all three illustrate our ongoing commitment to updating and enhancing our solutions over time. The breakthroughs described in our patents swiftly find their way into real-world deployment, and customers who use the SparkPredict product are getting cutting-edge technology (and in consequence, cutting-edge business benefits).
If you think capabilities like these are extraordinary, just wait; you might be in store for even more remarkable breakthroughs in the years to come. While these patents may sound impressive, they really only scratch the surface of the ongoing innovation under development at the company every day.
Based on the established history, and looking forward to what is likely to be another record-breaking year in patents filed, it’s more than fair to characterize SparkCognition as a true thought leader in the artificial intelligence space.
You might even say it’s patently obvious.
