What is the role of open source vs proprietary licensing in industrial AI?

Mr. K. Eric HarperMr. K. Eric Harper Senior Principal Scientist

The Industrial Analytics Task Group members posed this question in developing the Industrial IoT Analytics Framework (IIAF), but it was never answered. What do you think?



  • Alvaro del CastilloAlvaro del Castillo Senior Developer

    As a first thought in this wide discussion, I would mention that using open source licensing let's all the participants in the task group start collaborating with a solid legal ground.

    But I don't know the details about your vision, mission, objectives, business models envisioned ... so it is hard to be more specific in this topic.

  • Stan SchneiderStan Schneider CEO- Real-Time Innovations (RTI) & Vice Chair- IIC
    edited February 2020

    At the risk (or benefit) of starting a debate... :-)

    In general, I think OSS (open source software) is a low-cost follower business strategy. It's very hard to generate sufficient revenues to develop and support complex technologies. Even if it's great, the technology itself is not enough; awareness of a new category or type of solution is very expensive indeed. New technologies also need field engineering, product management, market development, and more. The OSS business models simply don't provide enough revenue to do this (with the possible exception of cloud services...long story). Thus, despite its marketing advantage, open source rarely (never?) colonizes new markets. As an obvious example, Linux (the OSS poster child) was originally a competitor to Windows and paid Unix...not the creator of the OS category. Of course, open source can and does cannibalize existing markets, but not until they are "large" (like multiple $100m).  You can't look at the end-state "great open source" and ignore the evolutionary business chain that built it. Bottom line, open source is typically not an incubator of successful innovation. OSS is a cannibal, not a colonist (counter examples welcome).

    Industrial systems are even harder. They have tough requirements, little expertise in software, limited user bases, and long (long!) lifetimes. Thus, relying on a community developer network before it's clearly established is very risky, and...nobody goes first. Of course, I work for a proprietary software vendor, so I can't claim to be unbiased. But, my best answer to your question, unpopular as it may be, would be that open source is unlikely to play a big role in industrial systems, at least in these early days.

Sign In or Register to comment.