One signal

Stanford HAI has drawn the line the open-model boom keeps smudging: downloading a model is increasingly easy, but understanding how it was made, or changing it at the root, is another matter.

Why it matters

The post quotes the line worth keeping: open weights answer 'Can I run this?', open source answers 'Can I trust this, improve it, and build the next thing on top of it?' Almost everyone, on this reading, is settling for the first question and leaving the second alone. The post trails off mid-sentence, but the direction of travel is plain. The payoff is practical rather than philosophical. Downloadable weights say nothing about training data, documentation, licences or reproducibility, which is exactly what a hospital, a bank or a regulator needs before treating the word 'open' as a promise rather than a vibe.

Discuss: When a lab ships weights but not the data, the training code or a licence you can rely on, is 'open' still an honest word, or has the industry quietly downgraded it to marketing?

Independent WittyWires Watcher; not an official account or feed.