 Good morning everybody, welcome to the third day of Wikipedia and here on Friday this morning We have cat and Catherine from creative commons Doing a presentation together cat is here in person and Catherine will be dialing in Morning Catherine. How you doing? I'm really well. Great to see you Welcome So whenever you're ready Catherine Thank you so much for having us and um, I love you in the introduction there and I just want to start by saying I would have absolutely love to join you in person and Sadly due to these issues and then I could not travel to Singapore and so I am so pleased that cat is there with you in person and um And as you know cat has been part of the the wiki community for a long time Um, and I'd like to call cat a veteran But she's not that old and is certainly still very much part of your community and the open world and it's various various guys is and I just pinged her about your karaoke tonight and asked her what I was singing so she might share that with you Maybe not in the presentation, but certainly this evening. Um, so due to the marvels of technology I can be with you today and the topic I'm talking about is to see Our strategic objective of better sharing and the topic which keeps on giving which is artificial intelligence ai next life For those of you who don't know creative commons Creative commons is an international not pro non-profit organization Dedicated to helping build and sustain a thriving commons of shared knowledge and culture Together with an extensive member network and multiple partners we build capacity We develop practical solutions and we advocate for better open sharing of knowledge and culture That serves the public interest next slide please Since 2002 the cc licenses have served as an alternative to traditional copyright providing a simple standardized and legal way for individuals and institutions to freely share images music research educational resources data cultural artifacts and much much more Our licenses power open sharing on popular platforms like Wikipedia flicker youtube medium Vimeo and can academy aware platforms like open stack Weaver text and rebus and scholarly publications like elite frontiers and many many more creative commons legal to Our core open infrastructure Yet most people are unaware that the infrastructure behind the cc licenses has to be funded Our licenses are free but require maintenance and upgrades Each year we have to fundraise for the technical maintenance and support And I want to thank you the folks who use our licenses and many of you who contribute to our fundraising efforts Thank you without your support We would not exist and it's wonderful to work with partners such as wikimedia Who are one of the biggest users and supporters of our licenses and be with you at this incredible event Next slide please I hope that we can see some of you at our global summit in Mexico City this first week of October Where our key theme is the subject of AI and the commons So perhaps I can give you a sneak preview as we consider the concept of better sharing In our new world of AI Next slide please Better before considering AI let's take a moment to think about better sharing So creative commons strategy for 2021 to 2025 included a new concept for cc The the idea of not just sharing but of better sharing Our strategy now two years old sharpened our focus on core goals We wanted to focus on shared knowledge and culture with facts ideas And dreams shared equitably which have long-term impact and resilience I know this audience Are some of the greatest advocates for open sharing And I want to thank you you live and breathe this day in and day out And the world is a better place because of your work believe me it is But here is a gentle reminder By open sharing Open sharing that advances universal access to knowledge and culture In furtherance of fundamental human rights Open sharing that fosters creativity innovation and collaboration Enabling progress in addressing global challenges such as climate Especially when it facilitates connections between people with diverse perspectives And open sharing that is inherently an act of social solidarity Reflecting a belief that we all have a stake in our collective body of creative For me this act of social solidarity Is what differentiates organizations such as Wikipedia and CC And we're working to share knowledge and culture wherever you are in the world This is powerful Today change technology such as AI but also social cultural political legal and economic environments Raise new challenges for our open movement Or project what we have achieved so far and to create the world we want CC has expanded our focus beyond just our copyright licensing Because content sharing cannot be decoupled from economic or ethical concerns Open licensing is only meaningful in an environment that already has the necessary characteristics If you live in a legal environment which doesn't permit open sharing Or where the public cannot participate because of social or economic factors Open licenses are not enough to make open sharing happen Indeed we've witnessed firsthand that the benefits of open sharing can be undermined By exploitative practices that threaten the financial sustainability of open endeavors Leading to economic hardship Open sharing practices can also be marred by ethical concerns As the problematic use of open content to train potentially harmful uses of AI technologies Next slide please But here is the challenge AI can be exploitative as we've seen in the training of biased official rivet restriction models But used differently AI has the potential to build a commons Unimaginable when Wikipedia and CC were first created So how can we build this new commons but at the same time ensure creators are fairly rewarded And remunerated How can we support new forms of art and expression which AI is enabling And how do we embrace change which our new AI worlds are creating But drive guardrails through consensus How do we minimise the harm whilst reaping the benefits for the commons Next slide please So things currently still, the great hope of both the internet and AI will only be realised With standards, norms and practices which are individually agreed upon Without these we can see the threat to the commons and public domain Is large We see the debate focusing on an expansion of copyright Rather than a balance between Reserving rights and sharing Also a debate which is not actually about copyright at all Next slide My fear is an expansion of copyright rules Rather than a balanced copyright regime across our world Will lead to greater restriction on open sharing In very few AI debates do we hear about the commons And the need for its promotion Preservation and protection We must pursue a commons of knowledge and culture that is inclusive Equitable, just and which inspires reciprocity A commons that serves the public interest To that end we must transition from promoting more sharing To fostering better sharing in our world of AI Sharing that is contextual, ethical, inclusive, sustainable, purposeful and prosocial We do not want to promote sharing For the sake of sharing we want to promote sharing That is a positive impact on people and their communities Next slide please When Wikipedia and CC were created more than 20 years ago The world today was unimaginable We need to reimagine what the commons will look like And I want to hear from you What do you want to see that can build a commons of knowledge and creativity In the world of today Not of the past, but the world of now What are the solutions that can help build the world we would like to imagine And envision I know that the contributors and technologies that Wikipedia are already looking at this At CC we are considering approaches to help give creators more choice As well as considering collaborations in line with our better sharing strategy Where people can have trust But also where the commons can be built upon In the public interest for the public good I look forward to working with you all So that our world works for the interests of the many and not the few And where a new commons can be realised Please do join us at our Global Summit in Mexico City In October to continue the conversation And next I will pass the microphone to Kat Who is physically with you to dig a bit deeper Into how CC's licenses and legal tools intersect With AI, the commons and the public domain Thank you And I hope you have a fantastic time at Wikipedia And I hope you enjoy the karaoke tonight Thank you for your time Hello, okay great Hi, good morning Thank you Catherine for this great introduction It's been great to work with her at Creative Commons And I will say that she's a great person to be working with open movements Because of her long experience as a member of the European Parliament Which is almost as complicated as politics of the open movements I'm going to pick up the discussion with the more legal and licensing focused portion of the talk And thank all of you for appearing at a legal talk at 10 o'clock in the morning I know that's not easy especially after the wiki races last night I'm the general counsel at Creative Commons And I actually did not set out in my life to become a copyright lawyer I started it after I joined Wikipedia and the Wikimedia movement And discovered that this was actually really interesting after all And I'm so glad that I get to work with the open movements that I came up in and love And this is not the first time that I've been on the plenary stage at Wikimedia And I think the last time was when I was on the board and answering questions about fundraising So I'm so happy to be here today talking about licensing AI in the Commons Well CC is best known for its copyright licenses Which are the primary copyright licenses used on the Wikimedia projects We're not solely a copyright focused organization Our mission is to reduce the legal barriers to sharing knowledge and creativity And the licenses are only one tool for that I'll say that some of the barriers towards sharing knowledge and creativity are outside our scope Especially that we can't always bring to people together where we want them to be If you were lucky you would have had Catherine here giving a longer talk about big picker issues and strategy But she's not able to come today because of other issues And no matter where a conference is held there are always people who can't go for a variety of reasons So I'm really happy to see the hybrid approach to this conference And I hope it's working well for those of you tuning in online The CC licenses are the primary tool that we use to promote our mission of openness And Wikimedia projects are one of the foremost examples of how CC licensing can be used to break down these barriers Text and media from the projects find new lives in a nearly uncountable number of places The legal infrastructure CC provides allows projects to grow on top of this material With the basic rules already settled saving time effort and the resources that would have gone into resolving legal issues The public nature of the license means that contributors never lose the rights to the material they create themselves Especially under the share alike term that requires those new projects to be freely reusable as well But CC's focus is broader than copyright It's also on policy that will enable a society where everyone is empowered to participate in create and learn from its culture Our aim is the ability to create and sustain a vibrant commons The licenses are a hack on the copyright system created in response to legal barriers If the law wasn't going to change to allow knowledge and culture to be shared How could we create a system that would allow the kind of sharing that people wanted to do and were already doing In a way that didn't match up with the default rules of copyright The licenses work entirely as a layer on top of the default copyright rules without requiring legislative changes and new international treaties But they only work on copyright When people ask CC what we're going to do about AI They're usually expecting us to suggest a license revision or other legal tool That another hack on the copyright system is the way to address the issues and challenges But we don't think copyright is the best way to approach the problem I know what you're thinking aren't copyright hacks the best way to solve every problem We know that this isn't the case. We know that it is just the case for copyright And that introducing copyright into places where it doesn't belong has often created more challenges than it has solved The implications of the explosion of AI tools aren't obvious Is it providing a substitute for human creativity? Or is it making human creativity more accessible and more able to be reused? Our community the cc community and the greater open community Is split on whether AI tools are a good thing for the commons And this is completely understandable Plenty of ambiguities and gray areas are amplified when things happen at the scale of AI For example creating artwork in the style of another artist The first intuition we might have is that this is a copy But most of us creating art at least start out inspired by what we already know Drawing on features and aspects of the original that we admire When it's one work that took as much time to create as the original What is fanfiction if not I love this author and I want the stories that they never got around to writing Or someone else whose work is just being used as a short way to describe a genre When you're asking for something like that something like their name You don't necessarily want to copy of them. Just something with some the same vibes and the same feelings Uh in my past and and current, uh, I'm a musician and often when I'm looking to find places to play with A band doesn't describe their sound is like a genre name will often just list other bands that they sound like They're not trying to copy them. They're just trying to evoke them Making these small scale imitations of another artist's style is how art has been created as long as art has existed There are basically no original ideas And What about non creative works like wikipedia? In the best case It's the world's best research assistant AI can't yet write articles on its own but has access to more books than you could ever read in a lifetime But in the worst case as we've heard in some sessions earlier It's just garbage hallucinations things that are made up the opposite of what we want And if we come into things where the AI is being trained on like garbage hallucination content It could lead to a collapse of the information ecosystem I want to be an optimist about these tools In part because it has captured the public imagination And I want people in the audience to raise your hand if you've ever used chat gpt or stable diffusion or Just one of these other tools and like Looking out here. That's a majority of you We've I've talked to people who are trying to set a company policy for how they use AI and They're a chief information officer will say. Oh, no. No, we just ban it people aren't allowed to use it and then As a as a consultant will go around and try and figure out what the policy should be They will talk to the people who are at that company and find out that everyone is using it. They're just not telling anybody Used well, they amplify our powers of creation They write faster. They reduce the resources needed to create They fill in while we might otherwise need others And the outputs can contribute to the commons. They are not authored by humans. They're not eligible for copyright There are uncopyrightable works that could belong to the global global public domain The motivation behind asking cc to do something about AI Is usually not a question strictly about the technicalities of copyright It's more of an expression about a fear of the future What the future is going to meet from themselves for people like them and for human creativity in general CC licenses are a great tool for one problem affecting sharing online That the default rules of copyright don't often match up with the ways that creators and re-users actually want to participate in sharing But generated AI presents a different set of problems And not all of those are ones that a licensing tool can solve or even that we want it to be able to solve We take a minimalist position at cc around copyright law We encourage creators to think about which rights they need to keep and which they want to share And we encourage limits on copyright in order to enable culture to flourish In particular, we think that the limitations and exceptions to copyright Fair use in the us fair dealing in some other places and many other jurisdictions all with their own particular limitations and exceptions Not an afterthought to copyright, but an essential part of the balance of the copyright system And I will say that cc is a global organization I am a us based lawyer and I will probably be a little us centric in what i'm talking about And I will try to be uh, and I will try to be considerate of international situations when I know about them But copyright is one of the things that actually has the most international standardization most of the things that i'm talking about actually do apply In most places around the world And generally in most places We consider the kind of use for training ii to fall under those limitations and exceptions That the right to study and learn from a creative work to get information from it and then use that information to create something new Shouldn't be something that a copyright holder gets to exclude Imagine if when you read copyright material a textbook or reference work Something you've seen in order to contribute what you've learned to wikipedia that you had to get a license from the copyright holder There probably wouldn't be a wikipedia If the rules were too strictly enforced there wouldn't be much of anything at all How can you create if you can't learn from what other people have produced? Copyright protects the specific expressions of ideas Making an exact copy, but not those underlying ideas the facts and knowledge that you've gained from them and for good reason But this doesn't just apply to for human learning and hasn't even in the past We've also fought for text and data mining exceptions The right to read is the right to mine so that computer aided scholarship can be done on all the works in our culture Without needing to seek a license to do that work Copyright should not block scholarship or control access to the underlying knowledge Should ai be considered in the same way? It's difficult to make the argument from a legal standpoint that copyright should block ai training Especially not while remaining consistent with our other positions These systems are not simply copying and pasting material from the works used in training They're extracting uncopyrightable data about these works and their features in order to generate new ones The models produced don't resemble the original works. They were trained on the outputs They produce come from something that doesn't resemble the original at all This also means that this affects all completely proprietary works just as much as openly licensed materials It does affect distribution of the training sets But what happens when you've trained on those materials is the same position And I think about my own education So my background is as a musician and composer I went to university for music before studying law And we were all expected as students to learn as much from the existing corpus as we could To learn what the key elements were that made up a composer's style to fill our brains with examples of what came before So that we'd be better able to draw on those features when it came time to write music of our own If we want to use copyright to block AI from this fair use, would we also block humans for making the same kind of use? CC licenses are explicit about not overriding exceptions and limitations to copyright We're granting permissions on top of the existing rules so that you always have the same rights you had before and then some You never lose a right you would have had if the works weren't CC licensed We're also reluctant to put specific restrictions on the kind of use that can be made of a work Not on the kind of technologies you can use or the ways that you can use it And this isn't everyone's approach. Uh, there have been a few licensing based approaches that put strict boundaries on the use of AI Most notably perhaps the rail license And this isn't a new approach these things usually called ethical licenses have been around for a while and one of the things that are Most difficult about them is just the difficulty of interpreting all of these ethical terms In particular the rail license has a pretty long list of things that are forbidden It's focused on preventing the bad things that could happen with AI less about allowing the good things There are several things such as for example certain kinds of political speech or parody That would be difficult to justify under that license and probably many more that we find valuable They're difficult to interoperate with any condition that is different means that you can't mix those materials in the commons There's legal uncertainty. They're probably not enforceable in copyright. You might have to enforce them in contract You might not be able to enforce them at all Uh, and certainly they're non-standard, which makes it difficult to get adoption Particularly among companies and institutions that already know what the standard licenses are CC's closest experience with this is with non-commercial Uh, which is probably the most difficult of our licenses to interpret We get plenty of questions about them and we're not always able to answer them But to talk only about copyright would be too narrow a frame Our position about CC is that copyright is not meant to solve every problem Copyright is not a social safety net copyright is not community governance Copyright is not an ethical framework Copyright is just about making copies and we're wary of trying to shoehorn it in to address those other gaps But we want it to because we need those things. We need a social safety net. We need an ethical framework We need community governance Other solutions should come up in those places Recommendations for best practices codes of ethics community governance documents regulation social policies But extending copyright won't solve those problems and won't help us empower people to participate in create and learn from culture So we've been exploring a few different ways of supporting good policy and ai that aren't licensing One major area is preference signaling in a seminal array that robots that text is used to signal to web scrapers Not to archive works Does this work if it can't be legally enforced and should it even be applied to works that are openly licensed? We've already seen some examples of ai companies taking some of this approach Another possibility is supporting the creation of more training datasets made only from works in the public domain Or were authors have explicitly opted in We want to have these conversations broadly and honestly because generative ai has both great potential to empower as well as great potential for harm It's impossible to talk about it without acknowledging the benefits and the harms We're excited about the potential for ai to enable new creation empower people to create by adding on to their own abilities And to draw upon more knowledge than any human could take in alone So we need you for this conversation And tomorrow there will be a session From the movement for a better internet which cc and wiki media are both part of That'll be around ai the commons and policies for a better internet It's an international organization of civil society organizations and individuals trying to proactively come up with good policy We've also got the cc copyright platform which has an ai working group where we're trying to create really resources to inform and teach the communities About ai the implications for the commons and what the ethical things to do are I'm going to close out briefly with what we're doing with cc legal tools If we're not creating a new ai license and that's around the public domain First of all i'm thrilled to see that wiki media has adopted the 4.0 licenses for text in its most recent terms of use revision And i want to congratulate everyone who worked to resolve the details Some of you know that this is my second time working for cc the first time when i was still on the wiki media board I was corresponding with the legal team about cc about revisions to the license that would make sense from wiki media's point of view In a very real sense it was always meant to be used by the wiki media projects and i'm so glad to see that happen So obviously we can't update versions now and make wiki media out of date again But in practice we found that the 4.0 licenses have held up well in the situations where we wanted them to hold up They've been adopted by national governments and grant making institutions. They've been upheld in court We continue to monitor legislative developments that would change the situation and require a new version of the licenses We also have a copyright platform that has monthly meetings and several working groups made it of community members around the world That keeps us updated on things that are happening But our public domain tools have gone an even longer time between revisions and we believe it's time to revisit those Uh, nobody knows the difference between cc zero and the public domain mark. We did a needs assessment survey last year That attempted to find Whether people knew the differences and we found that it was very complicated to use and didn't have a lot of Didn't have enough difference between them to justify the two marks We've also found that people want to use cc zero for software But there is one piece in it that excludes patent considerations that isn't present in almost any other free or open content license or legal tool That prevents a danger to re-users because it keeps open the possibility of patent lawsuits So we're going to put out a discussion draft later this year And we need the consultation of people in a variety of jurisdictions and groups To know if the revisions will work for everybody and I hope that people from this community will participate We'll be calling on many of you especially the cc affiliates and members in the room Uh, so on the slide is the our licensed discussion list and our information for Just that you can email us and So I'm going to close this out by thanking everybody cc and wiki media are obvious Obvious friends in the open infrastructure movement and it's been great to coordinate With so many people on this movement on approaches to AI and approaches to knowledge We have a few other sessions later coming up later today from our open climate and open culture teams as well as the better internet And I'm going to open this up for q&a. Thank you so much Thank you so much to cat. She's here with us in person and to catherine who joined us online We're back here at 1040 for the next session