AI, open source and the future: a conversation
A long-form conversation about what happens when the most capable software of a generation is developed partly in the open, and who gets to shape it.
About this conversation
Open-weight models, open datasets and open evaluation are changing who can build with AI and who can scrutinise it. This conversation covers the trade-offs honestly: safety, concentration of compute, the economics of releasing weights, and what “open” should mean when the artefact is a model rather than code.
Corrections
None to date. If you spot an error, see our corrections policy.