Twitter thread draft (TTW)

https://twitter.com/turingway/status/1513541747095314441

⚠️🚨NEW CHAPTER ALERT! 🚨⚠️
A new sub-chapter on Machine Learning Model Licenses is now live, co-written by @jen_gineered and @Carlos_MFerr!
👉 https://the-turing-way.netlify.app/reproducible-research/licensing/licensing-ml.html
👉 https://the-turing-way.netlify.app/reproducible-research/licensing/licensing-ml-case-studies.html

These sub-chapters sit in our Guide for Reproducibility, covering:
👉 Definitions of ML model licenses, and how they differ from data or source code licenses
👉 Examples of existing ML models and licenses
👉 Case studies of concerns raised through the license selection process

Attached w/ alt text:

Alt: Screenshot of Machine Learning Model Licenses chapter: featuring introduction and 'reproduction and propogation of ML models' section. Features navigation bar on left side with sub-chapters for Guide for Reproducible Research.

Alt: Screenshot of Machine Learning Model Licenses chapter, focused on "examples of ML models and their licenses" section. Table with types of licesnses and descriptions. Features navigation bar on left side with sub-chapters for Guide for Reproducible Research.

This chapter emerged out of a @turinginst event held earlier this year on @BigscienceW

and their approach to building LLMs in the open. The discussion with @TuringPubPol, @turingway, and AI Programme continued after the event with @Carlos_MFerr, AI Counsel at @HuggingFace.

Months later, the result is a new chapter in The Turing Way on ML Model Licenses!

Licenses are an emergent topic in the field of #machinelearning and #AI. We are excited to support the codification of these ongoing practices through shared, community-led documentation… and hope you'll join in on the conversation too!

Sharing plans (ALS):

• post twitter thread in Creative Commons slack
• Mozilla slack
• AI for Design slack

Twitter Thread (Jen - ALS edits)

[name=Anne] Adding in some edits here just for spacing. I tend to use https://getchirrapp.com/ for TTW tweets, figured it might be useful for you too for tweet spacing!

In June, @turinginst hosted an event on @BigscienceW

and their approach to building LLMs in the open. As part of that discussion with @TuringPubPol, @turingway, and AI Programme, we touched on the thorny issue of misuse of open source ML models.

The conversation continued after the event with @Carlos_MFerr, AI Counsel at @HuggingFace… and months later, the result is a new chapter in The Turing Way on ML Model Licenses!
https://the-turing-way.netlify.app/reproducible-research/licensing/licensing-ml.html

Through work on the @BigScienceLLM, Carlos & his team identified that the license offers an interesting opportunity to integrate ethics-informed restrictions to shape how a model is allowed to be used.

Rather than simply copy/pasting MIT or Apache 2.0 into a license.md, a license can be a dynamic artefact that embeds the creator and community values into the application of a model and its derivatives!

Our chapter on ML Model Licenses is inspired by decades of work in the AI world.
[name=Anne] Wouln't fit in the above sections - but this is a deeply powerful sentiment.

Particularly interesting for me was the evolution of licenses over the years, from fully public domain projects like YOLO and @scikit_learn to exclusive and commercial licensing in models like GPT-3 and DALL-E, and now to the first Responsible AI Licenses (licenses.ai) used for models like @StableDiffusion.

If anything, working on this has made us realize how much more there is to capture! From different practices of "open" within the AI world to effective mechanisms for license enforcement… stay tuned for more in the months to come.

A huge thank you to @Carlos_MFerr for collaborating on this, @aleesteele, @arronlacey, & @brainonsilicon for their thoughtful comments, and @drsarahlgibson for the last minute save to bring this chapter live!

Twitter Thread (Jen)

In June, @turinginst hosted an event on @BigscienceW

and their approach to building LLMs in the open. As part of that discussion with @TuringPubPol, @turingway, and AI Programme, we touched on the thorny issue of misuse of open source ML models. The conversation continued after the event with @Carlos_MFerr, AI Counsel at @HuggingFace… and months later, the result is a new chapter in The Turing Way on ML Model Licenses!
https://the-turing-way.netlify.app/reproducible-research/licensing/licensing-ml.html

Through work on the @BigScienceLLM, Carlos & his team identified that the license offers an interesting opportunity to integrate ethics-informed restrictions to shape how a model is allowed to be used. Rather than simply copy/pasting MIT or Apache 2.0 into a license.md, a license can be a dynamic artefact that embeds the creator and community values into the application of a model and its derivatives!

Our chapter on ML Model Licenses is inspired by decades of work in the AI world. Particularly interesting for me was the evolution of licenses over the years, from fully public domain projects like YOLO and @scikit_learn to exclusive and commercial licensing in models like GPT-3 and DALL-E, and now to the first Responsible AI Licenses (licenses.ai) used for models like @StableDiffusion.

If anything, working on this has made us realize how much more there is to capture! From different practices of "open" within the AI world to effective mechanisms for license enforcement… stay tuned for more in the months to come.

A huge thank you to @Carlos_MFerr for collaborating on this, @aleesteele, @arronlacey, & @brainonsilicon for their thoughtful comments, and @drsarahlgibson for the last minute save to bring this chapter live!

Screenshots

https://the-turing-way.netlify.app/reproducible-research/licensing/licensing-ml.html