by Tom Beyer
JNS ChatGPT Experiment
The Journal of Neurosurgery (JNS), a stalwart PubFactory client of many years, is running an interesting experiment right now. We have helped them create a UI to allow users to use a special version of ChatGPT to query aspects of an article. Information about the experiment can be found here:
The manuscript itself can be found here:
And a FAQ about the experiment can be found here:
It’s exciting for PubFactory to be used for such a forward looking technological integration. Come to the Virtual Series if you want to hear more.
LLM / ChatGPT Links of interest
I’ve also compiled a few interesting links about LLMs & ChatGPT that further delve into the wild and wacky world that is rapidly evolving around us. This time feels like the late 90’s in that things are changing fast and clearly there are going to be some big consequences but no one really knows how it’s going to shake out.
Large language models, explained with a minimum of math and jargon
One of the best explainer articles that I’ve found. Essentially LLM’s work by converting words to number vectors in a very high dimensional virtual space. This clusters like words and concepts in interesting ways (you can do vector math on them) and seemingly when the vectors and spaces get large enough new capabilities start emerging – and no one knows exactly what, how or why.
Catching up on the weird world of LLMs
This is an eye opening account of things you can do with LLM’s from someone who is busy experimenting with them. It seems we are only scratching the surface of what attaching these systems to other things will bring. There’s also a scary section at the end about how things can go very wrong. New favorite metaphor – this is alien tech and we’re just poking at it with a stick to see what happens. 🤯
The Lone Banana Problem.
Or, the new programming: “speaking” AI
I like this article because generative image systems (which use the same underlying technology as LLM’s) showcase in an easier to understand way some of the subtle biases and blind spots that these systems can have.
OpenAI scuttles AI-written text detector over ‘low rate of accuracy’
No, we don’t know how to detect AI written text yet…
SSP Sponsored Session
KGL has been growing so quickly that it is hard (even for us) to keep up with all the things we can do. Our sponsored session at SSP tried to cover some of this and includes an interesting case study from AAA:

Contact your project manager if you would like to explore some of the other capabilities KGL offers.