Open-weight piggybacks onto all the great press that open-source has but offers few of the advantages. Open-weight models are NOT fully customizable, and not transparent. They are the product of machine learning (the weights of a trained neural network), but not the whole thing, from soup to nuts. That means you can run them (or “posttrain” them), but, crucially, users can’t customize or improve them to remotely the same degree (more about that in a second), and outsiders can’t investigate them to nearly the same degree. You also can’t build community in the same way.
If pigeons are “just rats with better PR”, open-weights is open-source with all the good press but far fewer advantages.
No comments:
Post a Comment