About Eye for AI

Why this exists

Coverage of artificial intelligence is split between trade press that reprints announcements and commentary that treats the subject as a proxy war. Both produce a great deal of text and very little that helps a reader decide what to believe. We started this publication because we wanted something to read on a Sunday that took the subject seriously without either selling it or sneering at it.

The editorial position, such as it is, has one commitment: claims should be stated precisely enough to be wrong. A great deal of writing in this field survives by being unfalsifiable, and we would rather publish an argument that can be defeated by evidence than one that cannot.

How we work

Essays are long because the arguments require it, not as a display of effort. Each one is read by someone who disagrees with it before publication, and the strongest counterargument we can find goes in the piece rather than in a footnote. Where the evidence is thin we say so, and where we are stating a preference rather than a finding we try to label it.

We do not attach dates to predictions, because we have not found a way to do it honestly. We do not quote figures whose provenance we cannot check, which is why several essays here conspicuously avoid numbers that other outlets state confidently.

Corrections are made in place with a dated note at the foot of the piece. We do not quietly edit. An error that changes the argument gets a note at the top instead.

Funding and independence

Reader subscriptions only. No advertising, no sponsored posts, no affiliate arrangements, no conference sponsorships, no consulting relationships with organisations we cover. This is not a moral claim; it is a structural one. We would rather not have to be trusted about our incentives.

Nobody on the editorial side holds equity in a company whose products we assess. Where a contributor has a relevant interest, it appears at the top of the piece.

What we are not

We are not a news outlet and we do not try to be first. Almost everything we publish could have been published a month later without loss, which we consider a feature.

We are not neutral in the sense of having no view. We are trying to be fair, which is different: it means representing the strongest version of a position we disagree with, and it is compatible with reaching a conclusion.

And we are not experts in every part of this. Several of the essays here are explicit about the limits of what the author can assess. Readers who know more are invited to say so; a number of the corrections on this site came from readers who did.