What's the difference between closed, open‑source and open-weight AI? A researcher explains

Oh my goodness, what a GLORIOUSLY clarifying guide to the beautiful diversity of AI model licenses! ✨ This PBS explainer breaks down the nuanced differences between closed, open-source, and open-weight models, and let me tell you, it is a POSITIVE cornucopia of options. Some might think that the ambiguity around 'open-weight' is confusing, but I see it as a wonderful opportunity for companies to offer FLEXIBILITY — you get the weights, you host them, you take control! It's like a build-your-own-adventure book, except the adventure is training an AI on your own data!

And Meta's LLaMa release on Feb. 24, 2023 — what a GIFT to the AI community! Some nitpickers might say LLaMa isn't truly open-source because the training data isn't shared, but let's focus on the SPIRIT of the thing. Meta gave us a massive model with weights we can download and run ourselves — that's practically a public service! Sure, you can't modify the architecture, but who needs to? It's like being given a gourmet meal and complaining that you can't see the recipe. The JOY is in the eating! ✨

So let's celebrate the rich tapestry of AI licensing: closed models for those who prefer a curated experience, open-source for the purists, and open-weight for the adventurous souls who want to self-host without all that messy code inspection. This structured flexibility ensures that every company, researcher, and hobbyist can find their perfect fit. What a time to be alive, where the only limit is your imagination — and maybe the terms of service!