All insights

Quarterly review17 December 20241 min read

A common way to connect AI to everything else

AI researchers won Nobel Prizes, the Model Context Protocol offered an open standard for connecting models to data, and the year ended with a run of new releases.

What happened this quarter

In October, the Nobel Prize in Physics went to John Hopfield and Geoffrey Hinton for foundational work on neural networks. The Chemistry prize was shared by David Baker, Demis Hassabis and John Jumper, partly for AlphaFold, which predicts protein structures. Later that month, Anthropic released an upgraded Claude 3.5 Sonnet with an experimental ability to operate a computer through screenshots, clicks and typing.

In November, Anthropic published the Model Context Protocol (MCP), an open standard for connecting AI applications to data sources and tools. December brought OpenAI's full o1 release and Google's Gemini 2.0.

Our take

MCP is the announcement we expect to matter most to the organisations we work with. Until now, every connection between an AI tool and a business system was custom work, often tied to one vendor. An open protocol means a council could connect its document store or case system once, use it from several AI tools, and switch between them later.

That fits the Technology Code of Practice, which asks for open standards and no lock-in. We are adopting MCP for our own tools, starting with the document analysis we use on tender packs.

The Nobel prizes are a reminder of where much of the long-term value of AI may lie: in science, where it helps people answer questions that were out of reach.

Keep reading

Related insights