The Author’s Guild just added their own class-action lawsuit against OpenAI, claiming that using their copyrighted works to train ChatGPT violated their respective copyrights. This is essentially the same argument made in two other lawsuits filed a few months ago and in the class-action lawsuit filed by artists against Stability AI. As I said with […]
One of the big claims in the class-action lawsuit against Stability AI is that Stable Diffusion in some way contains all its training data, and is therefore a derivative work it its own right: Because a trained diffusion model can produce a copy of any of its Training Images—which could number in the billions—the diffusion
Well, the long-anticipated copyright battles over AI-generated content have finally started. Last week a group of artists announced they are suing Stability AI, Midjourney and DeviantArt for using their artwork (and that of literally millions of other artists) to train their machine learning systems, claiming doing so violates their copyrights. And yesterday Getty Images announced
The White House Office of Science and Technology Policy (OSTP) just announced that all publications and supporting data stemming from federally-funded research must be soon also be made available to all, without an embargo period or cost. The Open Access movement has made a lot of headway since Aaron Swartz‘s early activism, and since 2013
A couple years ago, the Australian quiz show “Spicks & Specs” asked its panelists to name the Australian folksong that could be heard in a popular hit single that was first released in 1979. The answer: “Kookaburra Sits in the Old Gum Tree,” in the flute riff of the Grammy-winning band Men At Work‘s hit single, “Down Under.”
That quiz show prompted Larrikin Publishing, who bought the copyright for the now 68-year-old folk song after its composer’s death in 1988, to sue for copyright. And yesterday a Sydney judge declared yesterday that the 11-note flute riff did indeed copy from the folk song, and will determine what royalties might be owed by the band.
Despite what some breathless news reports are claiming, damages will likely be limited — as CNN reports, the Larrikin is only claiming a percentage of revenues on Australian sales from the past six years, and the judge has already noted that he has not found that the flute riff is “a substantial part of Down Under or that it is the ‘hook’ of that song.” Still, it’s gotten me thinking about how many other copyright land mines might be out there, just waiting for someone (or some thing) to uncover the similarity between some riff and some other previous melody.
Musicians are always borrowing riffs and melodies from previous songs, from little riffs jazz musicians throw in as shout out to other songs to wholesale note-for-note copying. A few well-known examples include The Beach Boys hit “Surfin’ USA,” a note-for-note copy of Chuck Berry’s “Sweet Little Sixteen.” (Berry was granted writing credits to the former after a successful lawsuit.) The tune to the 1953 song Istanbul (Not Constantinople) is extremely similar to Irving Berlin’s Puttin’ On The Ritz. And the chorus to the 1923 hit “Yes! We Have No Bananas” is almost entirely made up of riffs from other songs.
That’s just a few examples that have come to people’s attention, but how many are out there that borrow from less obvious sources? How many are just waiting for a game show (or a new search engine) to copyright holders to a potential opportunity for some quick royalties? In the past few years it has become possible to search a music database for a recording by playing a snippit of a song or in some cases just by humming a melody. What is not yet possible is to automatically process an audio stream, tease out individual riffs and melody lines, and then find other earlier pieces that contain similar riffs and melody lines. But that kind of research is ongoing, and I have no doubt that it will be solved at some point. When that day comes, we will in essence be able to map out the genome of every music recording ever made, and from that we can lay bare the lineage of every song in history.
When that happens, how many other Kookaburras will we find?
The right to enforce patents is one of the powers specifically spelled out in the U.S. Constitution, which states that Congress shall have the power
To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries;
A new study published in The Columbia Science and Technology Law Review suggests that, in fact, patents deter innovation. The authors (one of whom, coincidentally, was my old roommate in grad school) created a patent simulation game that allows players to “invent” new products by arranging a sequence of widgets. These products can be sold to consumers, and the value of a sequence in the marketplace is related to its subsequence, so it makes sense for players to try to build off of particularly valuable sub-sequences.
Once a player has invented a previously undiscovered sequence, he may choose to open source the discovery or to pay a fee and patent it. Open sourcing a sequence simply prevents anyone from patenting any sequence based on it, while patenting a sequence allows the patent holder to license the sequence to other players and to sue anyone who infringes on the patent. If a patent holder decides to enforce his patent against an infringer, both players decide how many lawyers they wish to hire (again for a fee), and the case is decided by (virtual) die roll. Patent holders may also sell a patent outright to another player.
The researchers ran subjects in either a pure-patent version of the game that did not allow open source, a mixed version that allowed both patent and open source options, and a pure-commons version where patents were not allowed at all. Players were recruited from the incoming law school class, and were told that the player with the most money at the end of a trial would be given a prize. Their results show that players in the pure commons version produced more innovation (number of inventions), more productivity (number of inventions made) and higher social utility (amount of money each player ended with) than either of the other two variations. (The amount of innovations was not statistically significant, the other two metrics were very significant). Interestingly enough, they found no significant difference between the pure patent system and the mixed system for any of the three metrics.
It’s easy to nit-pick these kinds of simulation-based experiments, both in terms of how parameters are set and more generally whether the simulation captures enough of the real-world dynamics to be useful. One nit I have is that (near as I can tell) the market value of a product is the same regardless of how many competitors are selling the same product, which would eliminate one of the primary purposes of gaining patent protection. I also wonder whether the stated goal of making more money than your fellow players discouraged strategies that help everyone equally (a rising tide raising all ships), and in particular whether it might have discouraged use of the “open source” option in the mixed variation.
That said, it’s an interesting study, and in their discussion the authors cite many empirical and theoretical studies in the past few decades that have also brought into question whether patents actually promote innovation in the real world. The authors also suggest the possibility of more studies using their PatentSim game, and possibly even creating an online massive multiplayer version, which would presumably allow players to develop their strategy and experience with the game over longer periods of time.
From the Creative Commons weblog:
Creative Commons is excited to launch a beta version of its “Returning Authors Rights: Termination of Transfer” tool. The tool has been included in ccLabs — CC’s platform for demoing new tech tools. It’s a beta demo so it doesn’t produce any useable results at this stage. We have launched it to get your feedback.
Briefly, the U.S. Copyright Act gives creators a mechanism by which they can reclaim rights that they sold or licensed away many years ago. Often artists sign away their rights at the start of their careers when they lack sophisticated negotiating experience, access to good legal advice or any knowledge of the true value of their work so they face an unequal bargaining situation. The “termination of transfer” provisions are intended to give artists a way to rebalance the bargain, giving them a “second bite of the apple.” By allowing artists to reclaim their rights, the U.S. Congress hoped that authors could renegotiate old deals or negotiate new deals on stronger footing (and hopefully with greater remuneration too!!). A longer explanation of the purpose of the “termination of transfer” provisions is set out in this FAQ.
Basically their tool is designed to help authors and artists navigate the legal waters and reclaim their copyrights. From Lessig’s blog:
Why is this a Creative Commons project? We’ve seen CC from the start as a tool to help creators manage an insanely complicated copyright system. When we have this running, we’ll offer any copyright owner who has reclaimed his or her rights the opportunity to distribute the work under a CC license. But that will be optional. Right now, we’re just offering the tool to make it simpler for authors to get what the copyright system was intended to give them.
EFF has a call out for prior art to help bust two broad patents:
The Patent Busting Project fights back against bogus patents by filing requests for reexamination against the worst offenders. We’ve successfully pushed the Patent and Trademark Office to reexamine patents held by Clear Channel and Test.com, and now we need your help to bust a few more.
A company called NeoMedia has a patent on reading an ‘index’ (e.g, a bar code) off a product, matching it with information in a database, and then connecting to a remote computer (e.g., a website). In other words, NeoMedia claims to have invented the basic concept of any technology that could, say, scan a product on a supermarket shelf and then connect you to a price-comparison website. To bust this overly broad patent, we need to find prior art that describes a product made before 1995 that might be something like a UPC scanner, but which also connects the user to a remote computer or database. Take a look at the description and please forward it to anyone you know who might have special knowledge in this area. You can submit your tips here.
Also in our sights is a patent on personalized subdomains from Ideaflood. For example, a student named Alice might have personalized URL ‘http://alice.university.edu/’ that redirects to a personal directory at ‘http://www.university.edu/~alice/.’ Ideaflood says that it has a patent on a key mechanism that makes this possible. We need prior art that describes such a method being used before 1999, specifically using DNS wildcards, html frames, and virtual hosting. Prior art systems might have existed in foreign ISPs, universities, or other ISPs with web-hosting services. You can submit tips here.“
I’ll betcha there’s prior art in the augmented reality field that reads on the first patent, either from Steve Feiner’s group at Columbia or maybe even the stuff we were playing with at the Media Lab. (I’ll go rooting around once I meet a different deadline I’m spending my evenings on…)
9. The method of providing user interface displays in an image forming apparatus which is really a bogus claim included amongst real claims, and which should be removed before filing; wherein the claim is included to determine if the inventor actually read the claims and the inventor should instruct the attorneys to remove the claim.