Strange New LLMs

I spent part of the weekend catching up on the latest season of Star Trek: Strange New Worlds (#selfburn), and I’ve been thinking since about the re-imagining of the holodeck in the era of large language models.

It seems we (or at least I) have finally come to an understanding of how such a technology might work. Not in terms of the immersive experience, but in terms of working with a technology to co-tell a story with us.

In the latest imagining, the holodeck isn’t just a VR video game on steroids, with prescribed paths and designated outcomes. Instead, it seems this season’s writers looked at the technology and intended for us to think, “OMG, it’s an LLM!” Characters create a project with a book series (one well out of copyright); they prompt the model with specific language (providing context, a task, and a goal); and they understand that the program is adapting to their responses (with disastrous effects, of course).

As I’ve written here before, I’ve mostly been using generative AI tools like ChatGPT as, well, tools. But I have pondered the possibilities for play. I’m still getting my books from the library because there are too many human-told stories to keep up with, too many voices to appreciate in my lifetime. But…the idea of creating a book or short story perfectly crafted to your own taste and mood in a specific moment is intriguing, I think. Rather than reading book reviews to try to find what you want, creating the entertainment you want, just for yourself, is now an easily available option.

It was a “we live in the future” moment for me, and made me pause to consider if I should try to have a little more awe for this very, very flawed technology.

Don’t worry, though. I also spent a good portion of the weekend reading Empire of AI, which keeps things right in perspective (more to come on that soon).