-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Deploying to master from @ b16ad1e 🚀
- Loading branch information
0 parents
commit 4669fdf
Showing
141 changed files
with
3,617 additions
and
0 deletions.
There are no files selected for viewing
Large diffs are not rendered by default.
Oops, something went wrong.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,77 @@ | ||
<!DOCTYPE html> | ||
<!-- | ||
This file was rendered by Pollen. Don't edit this file directly. It will be overwritten when Pollen re-renders. | ||
--> | ||
|
||
<html lang="en"> | ||
<head> | ||
<!-- Global site tag (gtag.js) - Google Analytics --> | ||
<script async src="https://www.googletagmanager.com/gtag/js?id=G-ZBTLXWG3QD"></script> | ||
<script> | ||
window.dataLayer = window.dataLayer || []; | ||
function gtag(){dataLayer.push(arguments);} | ||
gtag('js', new Date()); | ||
|
||
gtag('config', 'G-ZBTLXWG3QD'); | ||
</script> | ||
<meta name="google-site-verification" content="ApapaNT3CEd0OdSE-X9Xy4xF3r_gjtWDR05XS6FANu4" /> | ||
<meta name="msvalidate.01" content="E6A615B4A274D4C956DDF0ED5959BD59" /> | ||
|
||
<meta name="twitter:card" content="summary_large_image" /> | ||
|
||
<meta name="twitter:site" content="@sanchom" /> | ||
<meta property="og:title" content="Copyright Throughout a Creative AI Pipeline" /> | ||
<meta property="og:description" content="An article I wrote about copyright in the inputs, intermediate products, and final output of AI programs." /> | ||
<meta name="description" content="An article I wrote about copyright in the inputs, intermediate products, and final output of AI programs." /> | ||
<meta property="og:image" content="https://sanchom.github.io/assets/cortes-wheat.jpg" /> | ||
|
||
<meta charset="UTF-8"> | ||
<meta name="viewport" content="width=device-width, initial-scale=1.0"> | ||
<title>Sancho McCann—Copyright Throughout a Creative AI Pipeline</title> | ||
<link rel="stylesheet" type="text/css" href="../site-style.css" /> | ||
<link rel="canonical" href="https://sanchom.github.io/AI-copyright.html" /> | ||
|
||
<link rel="icon" type="image/png" sizes="16x16" href="assets/favicon/favicon-16x16.png" /> | ||
<link rel="icon" type="image/png" sizes="32x32" href="assets/favicon/favicon-32x32.png" /> | ||
</head> | ||
<body > | ||
<header> | ||
<nav> | ||
<div class="header"> | ||
<div class="left-header">◄ <a href="atlas-of-ai.html"><em>Atlas of AI</em></a></div> | ||
<div class="center-header"><a href="index.html">Home</a> · <a href="site-index.html">Index</a></div> | ||
<div class="right-header"><a href="statutory-research.html">Statutory research</a> ►</div> | ||
</div> | ||
</nav> | ||
<div style="clear: both;"></div> | ||
|
||
</header> | ||
<div style="clear: both;"></div> | ||
<article> | ||
|
||
<h1>Copyright Throughout a Creative AI Pipeline</h1><div class="byline">By Sancho McCann · <span class="date"> | ||
<time datetime="2021-07-27">2021-07-27</time>, <a href="https://github.com/sanchom/sanchom.github.io/commits/master-source/AI-copyright.html.pm">edited</a>: <time datetime="2023-03-26">2023-03-26</time> | ||
</span></div> | ||
<root><figure><p><img potential-feature="potential-feature" src="assets/cortes-wheat.jpg"/></p><figcaption>This is an image I created using the AI tool, <em><a href="https://deepdreamgenerator.com">Deep Dream Generator</a></em>. I used it to apply the style of Van Gogh’s <em><a href="https://en.wikipedia.org/wiki/Wheat_Field_with_Cypresses">A Wheatfield with Cypresses</a></em> to a photo that I took of a bay at Cortes Island. That website does not claim ownership of the AI output.</figcaption></figure><p>My article, “Copyright Throughout a Creative AI Pipeline,”<span class="sidenote-wrapper"><span><label for="fn-1" class="margin-toggle sidenote-number"></label><input type="checkbox" id="fn-1" class="margin-toggle"/><input type="checkbox" id="fn-1-expand" class="margin-expand"/><label for="fn-1-expand" class="sidenote" hyphens="none"><span class="bibliography-entry full-form-citation" data-citation-id="AI-copyright.html.pm-McCann" data-citation-pinpoint="false" data-citation-parenthetical="false" data-citation-judge="false" data-citation-speaker="false" data-citation-signal="false" data-citation-terminal=".">Sancho McCann, “<a href="https://digitalcommons.schulichlaw.dal.ca/cjlt/vol19/iss1/5/">Copyright Throughout a Creative AI Pipeline</a>” (2021) 19 Can JL & Tech 109<span data-short-form-placeholder="AI-copyright.html.pm-McCann"></span>.</span></label></span></span> was just published by the Canadian Journal of Law & Technology. It is available open-access <a href="https://digitalcommons.schulichlaw.dal.ca/cjlt/vol19/iss1/5/">here</a>.</p><p>This work is increasingly relevant as AI tools such as Dall-E, Stable Diffusion, ChatGPT (and other large-language models—LLMs) are producing arguably novel outputs. And the question of who owns the copyright to the model weights or parameters has become relevant given the <a href="https://www.vice.com/en/article/xgwqgw/facebooks-powerful-large-language-model-leaks-online-4chan-llama">leak of the model parameters behind one instance of Facebook’s LLaMa</a> (Large Language Model Meta AI). One <a href="https://twitter.com/d_feldman/status/1631761126219292672">Twitter user asks</a>, “Is redistributing the LLaMa weights [] even legal? Can copyright cover a big table of machine generated numbers?” I hope this paper provides a starting point for thinking about these problems.</p><h3 hyphens="none">Abstract</h3><p>Consider the following fact pattern.</p><blockquote><p>Alex paints some original works on canvas and posts photos of them online. Becca downloads those images and uses them to train an AI (training configures the AI’s model parameters to useful values). Becca posts the resulting trained parameter values on her website under a license that reserves to Becca the right to use the parameters commercially. Cory uses those parameter values in a program that is designed to produce artwork. Cory clicks create and the program produces a work. This work is new to Cory, but it looks a lot like one of Alex’s original canvas images. Cory sells the work. Advise Cory about their potential copyright liability to Alex (for the substantially similar work that the program produced and that Cory subsequently sold) and to Becca (for taking Becca’s parameters and using them commercially, contrary to the license).</p><p>Cory clicks create again. The program produces another work, this time quite different from any of Alex’s original paintings. Cory shares new work on Instagram. Danny copies this image from Cory’s Instagram feed and sells a bunch of postcards that feature that image. Advise Danny about their copyright liability to Cory.</p></blockquote><p>These scenarios are not as contrived as they might initially seem. People frequently use copyrighted works when training an AI (more precisely: when training an AI’s parameters). The resulting trained parameters are being shared under licences that assume the parameters are the subject of copyright. People do use these parameters in programs that can produce novel content. The resulting work can be quite surprising to the end-user and there are generally no checks in place to ensure that the new works do not take too directly from the original training data. However, many of the new works will be quite different from any content already in the world. And the end-users of the creative program often claim copyright ownership over the resulting novel work.</p><p>I will first present the training and use of a creative program based on a neural network, a popular model that forms the basis of state-of-the-art creative AIs. Then, I will examine each of the issues just raised:</p><p>1. Does the person managing the automatic training of a neural network’s parameters obtain a copyright in the resulting trained parameters?</p><p>2. Does a person using a program that produces artistic output obtain a copyright in that output?</p><p>3. The automatic training of a neural network requires large amounts of example data (a training set). Can images from around the internet be copied for the purpose of training a neural network?</p><p>4. What if a person uses an AI to produce a work that looks substantially similar to one of the training examples? Is that an infringement? And who is infringing?</p><p>Today’s state-of-the-art “creative” AI tools are based on a technology (neural networks) that serve to separate the programmer and trainer from any of the eventual expression, even the expression stored in the automatically-learned network parameters. It would be very rare that a programmer or trainer might obtain copyright in the output from an automatically trained “creative” AI. However, there are a multitude of ways to use such an AI to produce output, many of which would very well justify awarding copyright to the end-user, especially when they use the AI as an elaborate brush with which to bring their own ideas to life in expression.</p><p>The current methods of training these creative AI tools requires large amounts of training data: existing works often protected by copyright. It is unclear whether Canada’s fair dealing user right allows for such copying for the purpose of training a neural network, particularly when not for private purposes. When a fair dealing user right is not available, this copying would be copyright infringement: unauthorized reproduction of existing works. Canada should clarify or expand the fair dealing user right to allow for such copying.</p><p>Trainers must be careful that they have not simply embedded a representation of the training examples in the AI. If the AI effectively contains “direct reflections” of the training data such that it regularly reproduces them, distributing such an AI would be copyright infringement. The trainer has a burden to verify that they are not distributing copies of the training data.</p><p>This analysis allocates copyright in a manner consistent with a pragmatic conception of creativity and art. It keeps the focus on human expression and allows for free distribution of the material needed for more people to have more practice with creative tools while preserving protection for original expression.</p><h3 hyphens="none">Acknowledgements</h3><p>I would like to thank Professor Jon Festinger, Q.C., for many helpful discussions while supervising this work and Professor Graham Reynolds for valuable feedback on an earlier draft.</p><div class="endnotes print-only"><p><h2 hyphens="none">Notes</h2></p><p class="footnote" id="fn-1">1. <a href="#fn-source-1" class="backlink undecorated"> ↑ </a><span class="bibliography-entry full-form-citation" data-citation-id="AI-copyright.html.pm-McCann" data-citation-pinpoint="false" data-citation-parenthetical="false" data-citation-judge="false" data-citation-speaker="false" data-citation-signal="false" data-citation-terminal=".">Sancho McCann, “<a href="https://digitalcommons.schulichlaw.dal.ca/cjlt/vol19/iss1/5/">Copyright Throughout a Creative AI Pipeline</a>” (2021) 19 Can JL & Tech 109<span data-short-form-placeholder="AI-copyright.html.pm-McCann"></span>.</span></p></div></root> | ||
</article> | ||
|
||
<div id="disqus_thread"></div> | ||
<script> | ||
|
||
|
||
var disqus_config = function () { | ||
this.page.url = "https://sanchom.github.io/AI-copyright.html" | ||
this.page.identifier = "AI-copyright.html" | ||
}; | ||
|
||
(function() { // DON'T EDIT BELOW THIS LINE | ||
var d = document, s = d.createElement('script'); | ||
s.src = 'https://sanchom.disqus.com/embed.js'; | ||
s.setAttribute('data-timestamp', +new Date()); | ||
(d.head || d.body).appendChild(s); | ||
})(); | ||
</script> | ||
<noscript>Please enable JavaScript to view the <a href="https://disqus.com/?ref_noscript">comments powered by Disqus.</a></noscript> | ||
|
||
</body> | ||
</html> |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,25 @@ | ||
This license applies only to the files with `.rkt`, `.p`, or `.pp` | ||
extensions in this repository. To be clear, it does not apply to the | ||
files with `.pm` extensions. | ||
|
||
MIT License | ||
|
||
Copyright (c) 2018--2019 Sancho McCann | ||
|
||
Permission is hereby granted, free of charge, to any person obtaining a copy | ||
of this software and associated documentation files (the "Software"), to deal | ||
in the Software without restriction, including without limitation the rights | ||
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | ||
copies of the Software, and to permit persons to whom the Software is | ||
furnished to do so, subject to the following conditions: | ||
|
||
The above copyright notice and this permission notice shall be included in all | ||
copies or substantial portions of the Software. | ||
|
||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | ||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | ||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | ||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | ||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | ||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | ||
SOFTWARE. |
Oops, something went wrong.