Cb Insights Captures the Irrational AI Exuberance in the Market
CB Insights, an analyst firm focused on startups, is holding its Aha! conference in early December where it will unveil its list of the top “AI 100” companies. That’s right, CB Insights had to select “just” 100 AI companies claiming to use AI in their product. (More on CB Insights and their conference in the December blog.) With so much market noise, it’s no wonder the Gartner Hype Cycle (another analyst firm, for IT audiences) has “Machine Learning” and “Deep Learning” at the highest level of their July 2017 Technology Hype Cycle report, in the “Peak of Inflated Expectations” category.
At the same time, CB Insights captures the irrational AI exuberance in the market with diagrams such as this one. This view, widely shared by industry observers, suggests the actual use of AI technologies is negatively correlated to the marketing hype. This is what we’ve seen in our experience.
For example, one vendor, who will remain anonymous, shared a case study claiming they used AI to solve a production issue. In the presentation, the vendor included the actual formula they used to find the correlation between a production process and product outcomes. Upon inspection, the formula was regression, which is a fine tool for that type of work, but claiming it is equivalent to AI? The formula they used wasn’t even a variation of regression, it was the most basic form of the machine learning algorithm, that has been a feature of Excel since its introduction in 2007 in the Data Analysis Tool Pack.
So a 10 year old Excel function – repackaged, relabeled, and hyped as AI. Certainly, the confusion between the flavors of cognitive computing are a factor here – AI, ML, DL (Deep Learning) and other variations of “Cognitive Computing” with the arguments about supervised and unsupervised variations partly to blame. It’s not like there is an AI police force to ensure what’s claimed is actually there under the covers.
With this as context – the hype, the cynics, and the irrational exuberance – it’s fair to wonder what is Seeq’s take on the market and use of advanced computing algorithms in Seeq applications. Our experiences in process manufacturing have certainly shaped our view and use of these technologies.
First, our focus is on end users – process engineers, analysts, and others – and on their ability to find insights in their data. Therefore, regardless of the technologies used in Seeq, the goal is always to make them accessible to the user who probably doesn’t have extensive data science expertise. Seeq empowers the person with the expertise, experience, and education to know what they are looking for and make that process faster. So as Seeq taps additional functionality to enable analytics it will implement these innovations in the context of easy to use, accessible features.
Second, we use whatever technologies we can to help our users and customers succeed. If that’s ML functionality, great, but we aren’t restricted to “only” ML-based features for the sake of market hype. There are a tremendous number of algorithms and innovations that Seeq uses to assist users who need to find insights in their data – digital signal processing for data cleansing, algorithms for detecting shapes in process signals, and map-reduce models for distributed computing – to name just a few. The “purity” of these algorithms as ML, AI, or otherwise, isn’t the point, the point is they are the right way to accelerate our customers to success.
Third, the algorithm isn’t the everything. Machine learning and other algorithms are an important part of the solution for analytics in process manufacturing and IIoT solutions, but they are only a part of the solution. There are any number of articles about the time required for “data wrangling,” which is the required data connectivity, cleansing, and contextualization to ready data for use. Seeq’s focus includes these preparatory steps to applying data analytics, which means from data connection to insight to distribution, Seeq will accelerate your experience.
Finally, Seeq enables users to expand and extend your analytics to whatever level is required. While we have yet to be stymied in a customer deployment due to the lack of a specific algorithm or function, we anticipate that customers, and data scientists in particular, will keep expanding their use of Seeq, and consequently their need for specific algorithms. Therefore, Seeq enables extensibility to additional algorithms through its REST API (for example to Python algorithms), OData (for integration with Azure Machine Learning Studio), and the integration of algorithms directly into the Seeq user experience. With Seeq there won’t be any ceiling to your efforts, regardless of your exploration requirements.
What is today hype and promises will over time mature to an expected and stable component of our work environment. By incorporating AI, ML, and other technologies, Seeq will facilitate your data analytics experience every step of the way.