Siemens Digital Industries Software and Tangent Works have partnered to incorporate Tangent Works’ artificial intelligence (AI) technology, InstantML, into Siemens' MindSphere, an Internet of Things (IoT) software as a service (SaaS). This enables their customers to instantly use Mindsphere’s data analytics without resource-intensive model training and management.
MindSphere uses IoT to derive operational data and enables industrial customers to make decisions based on factual information. The information provided by IoT devices, however, is not directly legible and translatable into action plans and models right away. Data scientists and engineers have to train their models to present the data in such a way that it is usable by managers and decision-makers.
Traditionally, it takes weeks, from IoT launch, to have a minimally trained model that provides the information in a usable way and training the model requires a skilled data scientist or engineer to help provide the correct pathways and learning outcomes. Thus, companies that hope to utilize IoT data with the help of AI and machine learning (ML) need to allow for additional time to receive results, while also paying resident data scientists, or consultants, to do their model training for them.
Siemens' new partnership with Tangent Works hopes to apply AI/ML decision-making sooner rather than later. Tangent Works is an international company founded on the premise of democratizing ML. As previously mentioned, machine models typically take time to be trained and require a data scientist either on staff or in contract to provide that training. Not every company that wishes to leverage IoT and AI/ML have the resource to devote to a data scientist and there aren’t enough data scientists in the world to meet the demands of a broad range of industries for which accurate and timely forecasts are essential to their competitiveness.
“Already before the pandemic, our research showed that a shortage of digital talent, especially in IT and data science, was the biggest challenge for companies looking to realize IoT, Industry 4.0 and AI initiatives … This challenge will get worse in the coming years and most firms will simply not find the experts that can build the machine learning models and integrate them into everyday operations,” said Knud Lasse Luth, CEO at IT Analytics.
This means that data scientists are more in demand than ever before and demand will continue to grow as the population of available data scientists remains relatively stable. This data scientist shortage increases the costs involved in getting AI/ML models trained the traditional way.
Tangent Works InstantML addresses this gap by replacing, or at the very least supplementing, a lot of the work done by data scientists and engineers. It offers a low-code solution that can be deployed and operated by employees that would otherwise be considered non-technical. These ‘citizen developers’ as Tangent Works envisions them would have the capability of assisting model training. But, they would not need in-depth theory or understanding of the underlying mathematics to produce the results their companies are looking to get out of AI/ML models. They will simply use the software, which will provide additional models and guidance as they make any specific adjustments and perform interpretation.
InstantML addresses the problem of data scientist scarcity by analyzing historical time-series data and including all the possible influencers. It figures out which data is relevant in a specific combination, over a time interval, to predict the output.
From Tangent Works’ vision, an architecture based on a field of mathematics known as information geometry, the tangent information modeler (TIM) was born. InstantML is an extension of the modeler’s automatic model building engine. It automates the forecasting and anomaly detection process. It also brings together engineering, model building and model deployment under one offering.
Tangent Works’ tangent information modeler (TIM) automatically and instantaneously generates models, and the associated metadata used to derive each model.
In a typical AI/ML deployment, data scientists handcraft models to fit their company’s analysis needs. This process is labor-intensive as it requires many hours of data cleanup, preparation, model building, deployment, as well as rounds of re-deployment and fine-tuning before models are launched and provided information that meets operational needs.
There are Automatic Machine Learning (AutoML) techniques and software available on the market that tries and addresses the problem with brute force solutions. However, there is still a need for manual feature engineering steps whereby data scientists or engineers manipulate and adjust the models before they can be deployed. InstantML does all of the adjustments and model training in one step, which allows for fast deployment—if there is historical performance and IoT data.
Using Tangent Works’ InstantML as an embedded part of Siemens' MindSphere gives industrial users the capacity to quickly add AI/ML capabilities to the applications and solutions they are already using within MindSphere. In other words, current and future data can now be quickly read cost-effectively, without the need for model training and leveraged to create actionable insights and prescriptive tools to help identify cases of predictive maintenance and possible performance optimization options.
By removing this barrier to entry, Siemens and Tangent Works are democratizing AI/ML and allowing for a wider variety of applications, including those for which traditional model training would have been too costly.
The partnership of Siemens and Tangent Works not only bolsters MindSphere’s current quality and resource management capabilities, but it also expands the availability of analytics far beyond what may have been originally envisioned. Now, many different models focusing on different predictive analytics can be brought to the forefront. This means that many insights will be discovered because of the low entry cost to produce this information.