Please note that this page does not contain legal advice. Rather, it is a set of guidelines to follow to make sure that your submission to the Summer of Math Exposition abides by proper copyright law and is eligible for winning the cash prize at the end.
In general, as long as you double-check to make sure you have permission to use all material (music, video, imagery, text, etc.) contained in your submission to the Summer of Math Exposition, you should be good to go. That said, copyright law is incredibly confusing, so this page should (hopefully) help you navigate a bit.
On Copyrighted material and Fair Use
In general, you are not allowed to use copyrighted material in the Summer of Math Exposition without permission; however, you can still use the content in such a way that it falls under fair.
What does this mean? Well…
Quick overview of Fair Use
Fair use is a bit of a gray area legally, but it is generally a way for creators to use copyrighted material in their own work by transforming its original intent. For example, if you make a video discussing the differences between two of your favorite animated shows, this could be considered Fair Use because your video is not merely providing the same function as those shows, entertainment, it’s serving a distinct function, namely an in-depth analysis. It’s even more likely to be considered Fair Use if you only show short clips of the content and not the content as a whole. In contrast, if you simply re-upload the animated shows on your youtube channel, that would be copyright infringement.
- Let’s say you are making a video on Iterated Function Systems and you found some images of fractals you can generate with the method. It’s totally fine to use those images as long as you are using them as examples to better explain your own points.
Totally not OK:
- Let’s say you are making a video and want some background music from Taylor Swift. Don’t do that.
- Let’s say you are making a blog post and find a great description of your topic in a textbook so you copy and paste that into your own work. That’s plagiarism. Don’t do that unless you properly cite the source and build off of it.
If you want to use copyrighted material in your video or blog post, err on the side of caution and don’t use it unless you either have the consent of the copyright holder, or your use fits the criteria of Fair Use. Otherwise, you might have a content claim to deal with, which is a huge pain.
For the Summer of Math Exposition, just know that we will not consider entries which violate copyright law.
On using specific software
It is somewhat tricky to determine whether specific tools can be used for the Summer of Math Exposition and it largely depends on individual usage policies set by the individuals and companies developing the software, itself. The core problem is related to the commercial usage of specific intellectual property (the software). Simply put: if you use a piece of software to make a youtube video and then proceed to put advertisements on that video, you are using the software to make money, which constitutes commercial use. In addition, the Summer of Math Exposition provides a cash prize to several of the top entries, which means even non-video entries constitute commercial use.
So let’s break this down. Here is a list of software you can and cannot use for this competition: Definitely OK:
- Any open source programming language (C, Python, Julia, etc). If some company somewhere is using the language to make money, then it stands to reason that you can as well.
- Virtually any other open source software with a traditional license: MIT, BSD, Apache, GNU, etc). Note that if you use the GNU license, you must also license any derivative software as open source as well, but it is still permissible to use such code commercially.
- Software with restrictive licenses (forbidding commercial use) that you have a formal agreement with or allows for commercial use. Note that we have explicit permission to use the following tools for this competition:
Definitely not OK:
- Unlicensed software. This means the code is under copyright of the original author and cannot be used without some sort of copyright agreement.
- Software with restrictive licensing (forbidding commercial use) without an agreement. Please ask the developer or ask us to contact a company for you.
Note that it’s possible to use copyrighted software under fair use if you use the code as a point of criticism or parody in some way. For example, if your video is analyzing different strategies to do math visualization, you can use code with restrictive licenses to compare it to other codebases.
Ultimately, just be careful and check the licenses of software before you use it.
On Creative Commons and Public Domain material
In general, you are permitted to use any material licensed under Creative Commons or Public Domain; however, there are some restrictions to keep in mind.
Creative Commons (CC) licensing is a bit confusing, but is generally for material that is meant to be reused in other copyrighted work. That said, the creator of CC work is still the copyright holder and still has rights, they are simply sharing some of those rights with others to re-use or distribute their work as well.
Let’s talk about what this means
CC0: Public Domain
Any content used under this license is fair game, no strings attached. That said, it would be nice to credit the creator of the material somewhere.
CC-BY: Creative Commons by an author
This is (probably) the most common CC license and the only one allowed to choose by default on youtube (more information on that in the “How to license CC work on youtube” section below). This license simply means you need to provide attribution to the original content creator when using CC-BY work. A good guide on how to do this can be found here
CC-NC: Creative Commons Non-Commercial
This license simply means that you cannot use the CC content in any content that generates revenue. If you are using it for a classroom, that’s totally fine; however, you cannot put it in a youtube video if you also have ads enabled because you could be making income from the ads.
For this reason, CC-NC can technically be used in the competition, but anything using CC-NC content is ineligible for winning the cash prize at the end. Such content can still be judged in the Peer Review and be manually judged by Grant, James, and the rest of the panel.
CC-SA: Creative Commons Share Alike
This license means that any derivative work should be under the same license. So if you use content under CC-SA, your future content must also be under CC-SA!
At first, this might seem like an unusual request, but it’s a simple way to protect the copyright holder and also ensure there is more open content for other people to use in the future. If people using CC-SA work were allowed to change the license, they could (in principle), change the content into CC-BY content and then use the work under copyright, ultimately profiting from someone else’s copyright!
CC-ND: Creative Commons No Derivatives
This means that if you want to use the CC content in your own work, you must share it exactly as it was created, no modifications allowed.
In general, treat CC-ND content as if it were under copyright. It can still be used under fair use without modifying the copyright of your final submission.
Putting it all together
CC licenses can also be mixed and matched, so if you see CC-BY-NC-SA, this means you need to abide by all the licenses:
- BY: provide attribution
- NC: you must use the content in a non-commercial way
- SA: you must use the same license on your own content (CC-BY-NC-SA in this case)
How to license CC work on youtube
For blog posts, it is clear how to license your content under CC: just put the CC license somewhere (preferably on every page). For youtube, things are a bit different.
Youtube does allow content creators to use the CC-BY license by default, but any other CC licenses are not explicitly allowed. For the purposes of this competition, if you want to use other CC licenses, use the CC-BY license for your video and then specify the license you would like to use in the description.
Note (again) that this is not legal advice and we are unsure of whether such licensing would hold up in court.
On AI-generated materialIn general, treat AI-generated material as if it were copyrighted by another party.
This means that generative AI models (such as ChatGPT, Dalle, Midjourney, StableDiffusion, etc) should not be used to create any text, graphics, audio, or other media used in the final submission for the Summer of Math Exposition; however, such models can be used in any other stage of the creative process. The output of AI models can also be used if it’s in a way that falls under Fair Use.
Note that there are no restrictions on using AI voice banks or other tools where every person willingly contributed to the training data. For this contest, we are making a distinction between general AI and generative AI — those that use some form of large language model, stable diffusion, etc. P This is a slightly nuanced issue, so we will provide our justification for this statement here, along with examples of appropriate usage.
The main reason we forbid the usage of generative AI models in the final submissions for the Summer of Math Exposition is a logistical one: we cannot provide a monetary prize to someone who does not fully own the copyright to their work.
Under current US law, it is unclear what license media from generated AI models should fall under. This ambiguity is the reason we wish to avoid generative models for this competition. For example, the output could be considered Public Domain (CC0) because it was created by something other than a human  or derivative work from its training data . In neither of these cases can the person using the model claim copyright of its output unless the model clearly output’s the user’s “mental conception” (which might be a surprisingly high bar to clear) .
On the one hand, we feel that generative AI models will be indispensable for creative individuals in all disciplines in the near future and feel that it would be against the spirit of the Summer of Math Exposition to discourage participants from learning tools that can help them create better content in the future. On the other hand, it would be irresponsible of us to allow people to potentially use software illegally in this competition. Once more laws are on the books, we will have a clearer idea of how AI can be used in content creation. For now, because it’s a gray area, we are dancing around it for this competition.
We are happy to discuss this further on discord
Examples of appropriate usage
Though we ask all participants to avoid using AI in their entries (and also flag any AI-generated entries in the Peer Review), there are a few ways in which AI can be unambiguously used in SoME3. Namely, if AI is used at any part of the creative process besides in the final submission or if the AI work is being used as in Fair Use.
Using it as part of a creative workflow
It is totally fine to use AI as part of a creative workflow if no AI-generated media is present in the final submission for SoME3.
Here are some specific examples:
- Let’s say you are making a video on the Hausdorff distance and are unsure of whether your script adequately explains the concept. You then ask ChatGPT for the key take-aways of your script to male sure your key-points are coming across. This is totally OK because you are using ChatGPT as a “second opinion” on the topic.
- Let’s say you are writing a blog post, but are unsure of what visuals to use in this case. You then ask Dalle to sketch up some ideas, which you then use as inspiration for your own visuals later on. This is also totally OK as long as you do not use the Dalle output in your final post. If you do decide to use the Dalle visuals in your post, then make sure you are using them in Fair Use (outlined below)
Using it in Fair Use
Let’s apply the Fair Use guidelines from the Copyrighted Material section above for specific cases of AI-generated media you would like to use for SoME3.
Using an AI-generated image in your video by comparing it to hand-drawn art or using it as justification for a broader narrative such as supplementary material to explain some mathematical concept.
Specific example: let’s say you are making a video on eigenvalues and you ask Dalle to generate a “generic human face” and compare that to a paper you found on eigenfaces. This is totally fine because you are comparing and contrasting Dalle’s output to a known concept in computer graphics as a way to better explain eigenvalues
Using ChatGPT to generate a paragraph in your script and explaining any inaccuracies you might have found.
Specific example: let’s say you are writing a blog post on determinants and you ask ChatGPT “What is a determinant.” You then incorporate this into your blog post as a way of showcasing what “most people” think about determinants and how your blog is going to change their perspective. This is totally fine because you are expressly using ChatGPT’s output in a broader narrative.
Totally not OK:
Using an AI-generated image in your video as “visual flair” to make a section of your video more visually interesting.
Specific example: You have a few seconds in your video where you are just “talking to the camera” without anything visual on screen and are afraid people are going to click off because of it. To fix this, you ask Dalle to generate an image on your topic and show that instead. This is totally not OK because you are not addressing the image or using it in a broader narrative.
Using ChatGPT to generate a paragraph in your script and pretending that you wrote it.
Specific example: You do not know how to write the introduction to your blog post or video and ask ChatGPT to do it for you. This is totally not OK because you are pretending the output of ChatGPT is your own work without delineating it from your other content.
So here’s a general rule of thumb: If you are using AI and your audience cannot tell what part of your submission was made by AI, then you are not using it under fair use.
- AI in the Public Domain: This argument follows the long-standing tradition of the US copyright office not to provide copyright to a piece of art unless it was created by a human. In such cases, the work falls under Public Domain. If this holds up for AI art, it would be possible to use it in videos, blogs, etc, for the purposes of this competition. Things to consider:
There is a very interesting discussion about why copyright cannot be extended to non-human entities with the Monkey selfie copyright dispute where a nature photographer managed to get a few monkeys to take a selfie of itself but neither the monkey nor the photographer were granted copyright of the image(s), and instead they were immediately put in the public domain. Also (because it’s interesting), the court case is Naruto vs Slater
It is not particularly likely that AI generated art will always be in the public domain. It only falls there for now while we wait on actual legislation.
Here is an interview with an attorney supporting this opinion (but note that it’s not binding legal advice)
- AI as derivative work: This argument follows the rights of each copyright holder in the
training data. If their work was used to create derivative work, then their copyright was violated; however, the key question is what constitutes “derivative work,” which is currently unclear. In most cases, the AI must generate work that is “substantially similar” to pre-existing art – not just in a similar style! This is the most interesting area to watch regarding the future
of AI in art. Things to consider:
There are quite a few lawsuits currently being processed regarding the rights of artists who have had their art used in generative AI models without their consent
https://githubcopilotlitigation.com/: This is a class-action lawsuit against Microsoft, Github, and OpenAI claiming that Github Copilot has violated the copyright of each user’s software repository.
GettyImages has sued Stability AI twice, in the US and in the UK, for using their images in their AI models. Note that they can make the claim that the AI output is “substantially similar” because the GettyImages watermark regularly appears on generated images
Individual artists have also come forward to file lawsuits against several companies
Though AI has been trained on copyrighted data, it might be considered fair use as in the case of The Author’s Guild v. Google. Because the scanned data was not made publicly available.
- AI when creating art that is the “mental conception” of the user: As of March 15th, 2023 the US Copyright Office has stated that some AI generated work may be copyrighted if (and
only if) the output from the AI reflects the author's own "mental conception." This was further enforced on May 11th with a new policy statement. Things to consider:
Note that the US Copyright Office rejected the copyright of an AI-generated comic book, but accepted the copyright of the text and arrangement of images.
It is unclear how much a human must interact with the generative models to claim copyright.
For this competition, we simply note that there is a high bar to clear if the user of an AI model wants to copyright its output.