The O’Reilly approach to generative AI

More than 50% of O’Reilly members currently use our platform to learn in the flow of work—quickly getting the answers they need to solve problems and move on. By putting the right tools at their disposal, we believe that percentage will grow. Exponentially.

One of those tools is generative AI, and over the last year we’ve seen how its capabilities have taken the world by storm. But we’ve also seen how large language models (LLMs) trained on publicly available information or content leveraged without permission can be fraught with problems—including copyright infringement. Certainly, human intervention is required to ensure reliability and accuracy, and prevent falsities from making their way into products and services.

But when used ethically, GenAI tools can augment jobs, solve large-scale problems, and create tangible economic opportunities. So, to ensure the content we distribute maintains the level of trust our readers and members have come to expect, O’Reilly has created a GenAI policy for our experts to follow when leveraging GenAI in the creation of their own works. In short, we expect authors and content providers to track their use of GenAI and share this info with their editors as they develop their content, just as they might with third-party content for permissions. Doing so provides the necessary human intervention to ensure reliability and accuracy.

We’ve also taken great strides to protect our experts’ copyrights—and give them credit and remuneration for their work—as we move into the next generation of learning tools.

In our quest to develop a reliable GenAI that enables learning in the flow of work, we’re training an LLM whose answers can be depended on because it’s being trained solely on trusted content—including the widely respected O’Reilly animal books and titles from nearly 200 trusted publishing partners including Pearson, Harvard Business Review, and Packt. We’re also training it with our expert-led live and on-demand courses, Superstream virtual tech conferences, audiobooks, and more. Anything we use to train our GenAI tool is content we stand by.

That content could be used to create sets of assessment questions at the end of a chapter to prove that a member has absorbed what they need to learn. Or summarize a four-hour conference so they learn faster. Or translate our original works into multiple languages so our content can be offered globally.

And it could deliver unique answers to tough tech questions instantly, pulling from a range of reputable sources to ensure accuracy and quality—so teams can learn in the flow of work even better.

At O’Reilly, our experts typically own the copyright to the books and content we publish and distribute. In fact, we were one of the first publishers to license authors’ works for publication—leaving ownership of their original ideas in their hands. We do retain the right to build derivative works from their content, but we believe our experts have the right to be aware of those plans.

That doesn’t change with the advent of GenAI on our learning platform; we will continue to respect our creators’ copyrights. Which is why we’re developing a GenAI tool that can track the sources of the answers it provides—and give credit and remuneration to authors and experts when it leverages their royalty bearing content.

All GenAI companies need to adopt ethical standards and figure out a way to pay content creators for their works used to train LLMs. By stepping forward and showing it’s possible, we hope to inspire more organizations to do right by both their users and their content creators.

Best,

Laura Baldwin
President, O’Reilly