Judge Delivers Blow to Artists AI Copyright Lawsuit

0:00Today on the AI Breakdown, we are doing an extended brief covering the latest in an artist lawsuit against AI image generation companies, Elon Musk's comments at the AI Safety Summit, and new medical research that shows AI is about twice as good as biopsies at identifying the aggressiveness of certain types of cancer. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube channel, our newsletter, and our Discord. Welcome back to the AI Breakdown Brief. In fact, today is an extended brief because it is the beginning of a vacation trip.

0:34And so I am recording it a day ahead with all of the most important events. We kick off today with a story around one of the many copyright-related lawsuits that is going on currently around generative AI. This particular lawsuit was filed by three artists, Sarah Anderson, Kelly McKernan, and Carla Ortiz, and targeted three platforms, Stable Diffusion, Midjourney, and DeviantArt, for using their works without authorization and in violation of copyright. Now, a judge has just made a preliminary ruling in the case based on a request to dismiss from the companies involved, and it's widely being reported as a blow for artists, although I don't think it's quite that clear cut.

1:11So the reason that it's being described as a blow to these artists is that the judge in the case, William H. Oreck, issued a 28-page decision that dismissed most of the claims the three artists had made. So there are a couple things going on in this particular decision. One is that the judge is making a distinction between works that have been officially and formally copyrighted versus works that have not. Now, in the US, a copyright is considered to exist, quote, from the moment the work is created. However, when it comes to bringing a lawsuit around copyright, those copyrights have to have been registered.

1:42So that's one issue and part of the reason that certain charges have been dismissed. As Artnet News describes, the judge also challenged the idea that all images generated by the model are somehow derivative of copyrighted works, if not all works in the training data set were registered. He noted that the artists will have to prove substantial similarity between a generated work and their image used to train the AI. Claims against DeviantArt and MidJourney were also dropped after the judge determined that the plaintiffs had failed to prove that those services had actually been involved in the scraping of images or AI training.

2:10But in the end, in most cases, the artists were given the ability to update their claims. Said a lawyer for the artists, The order issued yesterday by Judge Oreck is consistent with the views he expressed during the hearing on July 19th. Judge Oreck sustained the plaintiff's core claim pertaining to direct copyright infringement by Stability AI, so that claim is now on a path to trial. As is common in a complex case, Judge Orr granted the plaintiff's permission to amend most of their other claims. We're confident that we can address the court's concerns. We will be filing an amended complaint in November.

2:38In the meantime, discovery in the case is proceeding. So for now, quite clearly a mixed decision, and one we will continue to keep our eyes on. Next up, we head over to the UK where the AI Safety Summit is in full swing. As we have discussed, the summit got a major media boost at least when Elon Musk decided to attend, and sure enough, he was right away talking to reporters and making news. As the summit kicked off, he told Reuters reporter that the goal of the event was to establish what he called a third-party referee. He said, quote, What we're really aiming for here is to establish a framework for insight so there's at least a third-party referee, an independent referee, that can observe what leading AI companies are doing and at least sound the alarm if they have concerns.

3:16I don't know what necessarily the fair rules are, but you've got to start with insight before you do oversight. Elon also pushed back on concerns that the government was going to over-regulate before they understood anything. He said, I think there's a lot of concern among people in the AI field that the government will sort of jump the gun on rules before knowing what to do. I think that's unlikely to happen. Now, speaking of AI rules and governments and where these rules are coming from, in a very widely reported story, apparently the last straw to push President Joe Biden to actually signed this executive order, was watching a movie, Mission Impossible 7, Dead Reckoning.

3:52In an interview with Time Magazine, Deputy White House Chief of Staff Bruce Reed said that the president watching the Mission Impossible movie, if he hadn't already been concerned about what could go wrong with AI before that movie, he saw plenty more to worry about. Now, this is just a quote from a random guy for some color in a magazine article, so I will bite my tongue about the implications of a work of fiction inspiring a major and highly consequential policy, but, well, I'll just leave that there in the air. However, when it comes to U.S. policy, obviously the next step after the launch of the executive order was for Vice President Kamala Harris to head to the U.K.

4:28to participate in the AI Safety Summit. It seemed like Harris's goal was to, A, start to flesh out what these policies would actually look like, taking the skeletons from the executive order and adding meat to the bones, as well as to present a strong voice arguing, that short-term issues such as bias that shows up in LLMs were every bit as big a concern as the larger, more existential-type questions that the AI Safety Summit was focused on. One of the specific announcements, according to the New York Times, was that VP Harris was quote, set to announce that 30 other nations have joined a political declaration created by the United States that seeks to establish a set of norms for responsible development, deployment, and use of military AI capabilities.

5:05Meanwhile, as we will see in our episode tomorrow, The occurrence of the AI Safety Summit is dredging up some significant debates, and in particular, pitting the AI safety community on the one hand with the open-source advocate community on the other. Meta Chief AI Scientist Jan LeCun has been right at the center of that, but now Meta Executive Nick Clegg has also waded into the conversation, comparing the fervor and fear around AI to the moral panic in the 1980s over video games. He said, New technologies always lead to hype. They often lead to excessive zeal amongst the advocates and excessive pessimism amongst the critics.

5:39I remember the 1980s. There was this moral panic about video games. There were moral panics about radio, the bicycle, the internet. These predictions about what's going to happen next, what's going to happen just around the corner, often doesn't quite turn out as those who are most steeped in it believe. As The Guardian points out, Clegg's comments could presage just how difficult to fight this is going to be. He said, I'm still a sufficiently old-fashioned liberal to worry about the dead hand of the state. In this area, it is really important to allow innovators, builders, people who are ingenious in the way in which they ultimately entrepreneurially develop these technologies, to do so without immediately assuming that whatever they do next is going to pose some existential risk.

6:14Like I said, this is a fierce and increasingly caustic and contentious conversation, and we will get more into that in tomorrow's episode, so keep an eye out for that. Moving over to markets, AMD was the latest company that touches AI to report earnings, and the big focus was on how their new chip, which they are rolling out in the next quarter, might impact their competitive position vis-a-vis NVIDIA. Reuters reports that AMD on that call forecast around$2 billion worth of sales of AI chips next year, which helped buoy the company's stock after it also forecast fourth quarter revenue and gross margins below what Wall Street had estimated.

6:47Upon those initial estimates, the stock of the company fell as much as 4.6%, but then came up to roughly flat after talking about its projections for the MI300X chip. Another big company with AI-related announcements is LinkedIn. The company made two major announcements. First, that it had hit a billion members. And second, that it was releasing a new set of AI features for people who were looking for jobs. The company bills the chatbot as a job seeker coach and is being powered by OpenAI's GPT-4. So how might people use this chatbot? Well, when users are looking at a job posting, for example, they could select one of a few questions, such as, am I a good fit for this job?

7:22Or how can I best position myself for this job? Now, once they've made that selection, the AI bot could do things like analyze a user's LinkedIn profile and experience and make suggestions about what to emphasize in a potential application. The chatbot is also designed to point out potential gaps in their experience so that people can figure out how to address them. Now, given concerns around bias in AI when it comes to things like hiring or job applications, LinkedIn is going to pains to make sure that their models do not have some of those biased impulses. Another company that has announced a big AI update, or really actually launched a big AI update in this case, is Microsoft.

7:56Earlier this fall, Microsoft announced that it was bringing an AI co-pilot experience not just to specific applications within its Windows suite, but to the Windows operating system itself. Well, the company has now begun rolling out those new Windows 11 features, and the highlight is certainly this co-pilot AI assistant. Now, in many ways, the goal of this is to make Windows 11 the first major AI operating system. What does that mean? Well, one of the ways that you can look at the advance of generative AI is as a transformation and disruption when it comes to the fundamental interface for how we interact with computers.

8:27For many, many years, we've been taught a set of new types of interactions that are specific to telling computers what we want. Pointing and clicking, using a mouse to point and click on things we want to open, using specific commands that we have to learn in order to get the computer to do what we want. The advance of these co-pilots and AI assistants and chatbots has the potential to reimagine the entire computing experience to be powered by natural language conversation. It'll be an interesting test in bellwether to see how much users adopt this co-pilot AI assistant and for what when it comes to its rollout in such a big operating system like Windows 11.

9:02Now, speaking of chatbots, a trade group that represents newspapers, the News Media Alliance, announced research this week that suggests that AI chatbots are relying more heavily on news content than on some other types of internet content that they've been trained upon. From the New York Times, the News Media Alliance, a trade group that represents more than 2 ,200 publishers, including the New York Times, released research on Tuesday that said it showed that developers outweigh articles over generic online content to train the technology and that chatbots reproduce sections of some articles in their responses.

9:31The group argued that the findings showed that the AI companies violate copyright law. Now, if you've watched the interaction between the AI space and the news world, and if you've seen just how many news platforms have opted out of training, thanks to OpenAI announcing how companies could do that, it's no surprise to see this sort of argument and this sort of ratcheting up of the tension. Ultimately, again, as we started with on this episode, when it comes to these questions of copyright and AI, it's going to have to come down to court decisions, because that's the only way that things are going to get settled.

10:00Now, as we start to wrap up, we move over to the realm of medical research. Medical research and health in general is one of the areas that the people who are most optimistic about AI are most optimistic about. And certainly some new research seems to validate that opinion. According to a study by the Royal Marsden NHS Foundation and the Institute of Cancer Research, artificial intelligence was almost twice as accurate as a biopsy at judging the aggressiveness of some cancers. As The Guardian writes, The study found that the AI algorithm was far better than a biopsy at correctly grading the aggressiveness of sarcomas, a rare form of cancer that develops in the body's connective tissue such as fat, muscle, and nerves.

10:35So how would this actually help things? Well, by having a more accurate way of grading tumors and the aggressiveness of tumors, there are a number of positive outcomes. The most high-risk patients could be identified more quickly and treated more aggressively, while low-risk patients could be spared things like follow-up scans, extra hospital visits, and necessary treatments. Now, when it comes to the actual results of the study, the AI in this case accurately graded the aggressiveness of tumors 82 % of the time, while biopsies were only accurate in 44 % of the cases. They were also better able to differentiate between different types of sarcomas 84 % of the time, while radiologists were only able to tell the difference in 65 % of the time.

11:12Now, of course, this is just one study in one specific category of cancer, but it's part of a larger sense of how these tools are evolving and how much they could transform the health system as a whole. Lastly, in a preview of the type of cheesy end-of-year story that is just going to get louder and louder in the months to come as we get towards New Year's, Collins Dictionary has picked AI as its Word of the Year. The statement read, If computers were suddenly experts in that most human of domains language, what next? Cue an explosion of debate, scrutiny and prediction, and more than enough justification for Collins' 2023 Word of the Year, AI.

11:46In being named Word of the Year, AI beat out other contenders including Semaglutide, Canon Event, and Nepo Baby. Congrats to AI on your victory, and thanks to all of you for listening or watching. As I mentioned, tomorrow we'll be back with a long episode about this growing tension between the open-source AI advocates on the one hand and the AI safety folks on the other. But until then, peace.

12:16Thank you.