Frequently asked questions
We employ a rigorous and well-documented program management process to ensure engagements proceed smoothly and deliver outstanding results. Here are answers to common questions about the process.
What is the first step?
Complete our
self-assessment questionnaireand we’ll be happy to review the results with you through a virtual consultation.Our engineers can help clarify your use case and value proposition, and can walk you through our online demonstrator.
Click here to connect with us.
Is my plant ready?
Complete our
self-assessment questionnaireand we’ll be happy to review the results with you.The basic technical requirement is a rudimentary level of digitization. Specifically, you must be able to measure the important variables that affect your process (inputs), as well as the performance metrics of interest (outputs). These signals must be available through a networked PLC. In practice, we almost always find that customers’ production lines satisfy this requirement.
If you have not yet completed this basic level of digitization, our Solutions Team can recommend preferred hardware and local integration partners. Typically, basic digitization is not prohibitively expensive or time-consuming.
We also require an external connection for collecting data. We use this data to train your AI models on large computers in the cloud. Once trained, models can run “on the edge” using common, inexpensive hardware located inside your plant.
How much data do I need?
Surprisingly little. In many applications, sufficient data for training an AI controller can be collected in a matter of days.
Much of the “secret sauce” in our proprietary solution involves training useful AI models with a small amount of data. In fact, our core algorithms hold the world record in data efficiency for the type of machine learning we perform. Ask us for the evidence!
How long does it take?
After your plant has satisfied the items in our assessment checklist (see “Is my plant ready?” above), our Solutions Team can begin training and testing AI models for production in a short time - typically this requires days or weeks, not months.
The total length of your engagement will depend on the number of production lines and product types you wish to control, as well as the amount of customization you require for plant floor UI.
Is it secure?
Our solutions apply industry best practices for security and data management. In addition, our AI models incorporate proprietary low-level security and cannot be loaded or utilized without proper credentials.
We do require a networked PLC within the plant for communication with production equipment. Whether or not you use Liveline’s solutions, physical access to equipment controllers can be a threat vector if not managed properly. We will recommend best practices regarding physical access and other security measures inside your plant as appropriate.
What about ethical AI?
Ethical AI is an important societal topic, as Artificial Intelligence is used increasingly by public and private entities to make decisions that impact peoples’ lives.
Many researchers, practitioners, and observers have written extensively on the topic, such as the Harvard Business Review and C3.ai , a leader in Enterprise AI.
One issue concerns bias that can creep into AI model behavior due to choices made about which datasets to use for training. This bias may lead to adverse outcomes for certain groups of people - for example, when making automated decisions about consumer credit. There are many other issues to consider, such as the carbon footprint associated with energy consumed by computational resources.
Our applications of AI are strictly limited to analysis of sensor data from machines, and manipulation of machine settings in real time. Neither the data we collect nor the models we train can be used beyond that purpose. While this does not address all concerns about ethical AI, it provides insulation from common concerns about bias towards individuals.
Because of our specific domain of application, our AI does not meet the criteria of “High-Risk AI Systems” as defined in the European Union’s proposed AI Act of 2021, and therefore would not fall under its scope of regulation should the Act be adopted as EU law.