AI Confidential

It’s been a wild couple of weeks in AI news — and things keep accelerating.

Our Head of Digital Development Dan Stone shares his take on what’s happening—and what happens next:

A lot of public focus is going into OpenAI, Microsoft and Google around ChatGPT and Bard, and I’ve heard a lot of people say that power in this space will be concentrated there. But these last weeks have had a cascade of announcements about new Open Source (OSS) models, some backed by some interesting players.

Among these—StabilityAI, which also created the popular OSS image model Stable Diffusion—released “StableLM.HuggingFace released HuggingGPT, Databricks Released Dolly 2.0—notable not just for its OSS license but its availability for commercial use. 

A lot of focus is on ChatGPT, but in the long run, OSS models may in fact dominate in the same way OSS software currently largely dominates across the cloud industry. Seems like some folks at Google *might* share that perspective. The more this ecosystem grows the more we might start to see a shift in the way future tools are built and operate. In that vein — keep an eye on companies like MosaicML, which will train a large language model (LLM) for any company on its own data.

Last week, AI pioneer Geoffrey Hinton left Google specifically to warn of potential danger if we don’t tread carefully. There was also an interesting perspective piece with former Google AI Ethicist Blake Lemoine who was fired by Google last summer for speaking publicly about his belief that one of their internal models had achieved consciousness. Personally — he seems less out there to me now than he did when I first heard about his story, and I thought this interesting, particularly his description of what the last couple of years inside Google looked like in this space.