What to do about large language models
For the BMJ board, a talk on AI and the corporate response.
ChatGPT and so on are large language models (LLMs) — the of generative AI to do with text that exploded in November 2022.
Like any new technology, LLMs bring opportunities and risks. They can be used to automate whole workflows and get closer to consumers, but also to accelerate scams and to supercharge competitors.
And also, like any new technology, they’re hard to reason about. We don’t fully understand the their capabilities or the possibilities.
We’re in a capability overhang. So, as an organisation, what should you go?
In May 2023, I was invited to speak at the BMJ Board Strategy offsite.
(The BMJ runs the British Medical Journal and over 60 other scientific journals, and has a number of major products in the health and medical space.)
I broke down the areas of business disruption resulting from LLMs, giving examples and models to help think about them.
Then, having privately polled a number of other companies as to their own strategies, I put forward a range of options to seed the conversation — together with my own preferred approach: systemic, strategic pathfinding via prototyping and sharing.
This is a talk I had been developing before being approached by the BMJ, and I’m grateful for their assistance in refining and informing it. I’d be happy to give it again, or a version of it, to leadership groups in other organisations.