On Friday, Bloomberg reported that Reddit has signed a contract allowing an unnamed AI company to train its models on the site’s content, according to people familiar with the matter. The move comes as the social media platform nears the introduction of its initial public offering (IPO), which could happen as soon as next month.
Reddit initially revealed the deal, which is reported to be worth $60 million a year, earlier in 2024 to potential investors of an anticipated IPO, Bloomberg said. The Bloomberg source speculates that the contract could serve as a model for future agreements with other AI companies.
After an era where AI companies utilized AI training data without expressly seeking any rightsholder permission, some tech firms have more recently begun entering deals where some content used for training AI models similar to GPT-4 (which runs the paid version of ChatGPT) comes under license. In December, for example, OpenAI signed an agreement with German publisher Axel Springer (publisher of Politico and Business Insider) for access to its articles. Previously, OpenAI has struck deals with other organizations, including the Associated Press. Reportedly, OpenAI is also in licensing talks with CNN, Fox, and Time, among others.
In April 2023, Reddit founder and CEO Steve Huffman told The New York Times that it planned to charge AI companies for access to its almost two decades’ worth of human-generated content.
If the reported $60 million/year deal goes through, it’s quite possible that if you’ve ever posted on Reddit, some of that material may be used to train the next generation of AI models that create text, still pictures, and video. Even without the deal, experts have discovered in the past that Reddit has been a key source of training data for large language models and AI image generators.
While we don’t know if OpenAI is the company that signed the deal with Reddit, Bloomberg speculates that Reddit’s ability to tap into AI hype for additional revenue may boost the value of its IPO, which might be worth $5 billion. Despite drama last year, Bloomberg states that Reddit pulled in more than $800 million in revenue in 2023, growing about 20 percent over its 2022 numbers.
Advance Publications, which owns Ars Technica parent Condé Nast, is the largest shareholder of Reddit.
Enlarge/ Snapshots from three videos generated using OpenAI’s Sora.
On Thursday, OpenAI announced Sora, a text-to-video AI model that can generate 60-second-long photorealistic HD video from written descriptions. While it’s only a research preview that we have not tested, it reportedly creates synthetic video (but not audio yet) at a fidelity and consistency greater than any text-to-video model available at the moment. It’s also freaking people out.
“It was nice knowing you all. Please tell your grandchildren about my videos and the lengths we went to to actually record them,” wrote Wall Street Journal tech reporter Joanna Stern on X.
“This could be the ‘holy shit’ moment of AI,” wrote Tom Warren of The Verge.
“Every single one of these videos is AI-generated, and if this doesn’t concern you at least a little bit, nothing will,” tweeted YouTube tech journalist Marques Brownlee.
For future reference—since this type of panic will some day appear ridiculous—there’s a generation of people who grew up believing that photorealistic video must be created by cameras. When video was faked (say, for Hollywood films), it took a lot of time, money, and effort to do so, and the results weren’t perfect. That gave people a baseline level of comfort that what they were seeing remotely was likely to be true, or at least representative of some kind of underlying truth. Even when the kid jumped over the lava, there was at least a kid and a room.
The prompt that generated the video above: “A movie trailer featuring the adventures of the 30 year old space man wearing a red wool knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colors.“
Technology like Sora pulls the rug out from under that kind of media frame of reference. Very soon, every photorealistic video you see online could be 100 percent false in every way. Moreover, every historical video you see could also be false. How we confront that as a society and work around it while maintaining trust in remote communications is far beyond the scope of this article, but I tried my hand at offering some solutions back in 2020, when all of the tech we’re seeing now seemed like a distant fantasy to most people.
In that piece, I called the moment that truth and fiction in media become indistinguishable the “cultural singularity.” It appears that OpenAI is on track to bring that prediction to pass a bit sooner than we expected.
Prompt: Reflections in the window of a train traveling through the Tokyo suburbs.
OpenAI has found that, like other AI models that use the transformer architecture, Sora scales with available compute. Given far more powerful computers behind the scenes, AI video fidelity could improve considerably over time. In other words, this is the “worst” AI-generated video is ever going to look. There’s no synchronized sound yet, but that might be solved in future models.
How (we think) they pulled it off
AI video synthesis has progressed by leaps and bounds over the past two years. We first covered text-to-video models in September 2022 with Meta’s Make-A-Video. A month later, Google showed off Imagen Video. And just 11 months ago, an AI-generated version of Will Smith eating spaghetti went viral. In May of last year, what was previously considered to be the front-runner in the text-to-video space, Runway Gen-2, helped craft a fake beer commercial full of twisted monstrosities, generated in two-second increments. In earlier video-generation models, people pop in and out of reality with ease, limbs flow together like pasta, and physics doesn’t seem to matter.
Sora (which means “sky” in Japanese) appears to be something altogether different. It’s high-resolution (1920×1080), can generate video with temporal consistency (maintaining the same subject over time) that lasts up to 60 seconds, and appears to follow text prompts with a great deal of fidelity. So, how did OpenAI pull it off?
OpenAI doesn’t usually share insider technical details with the press, so we’re left to speculate based on theories from experts and information given to the public.
OpenAI says that Sora is a diffusion model, much like DALL-E 3 and Stable Diffusion. It generates a video by starting off with noise and “gradually transforms it by removing the noise over many steps,” the company explains. It “recognizes” objects and concepts listed in the written prompt and pulls them out of the noise, so to speak, until a coherent series of video frames emerge.
Sora is capable of generating videos all at once from a text prompt, extending existing videos, or generating videos from still images. It achieves temporal consistency by giving the model “foresight” of many frames at once, as OpenAI calls it, solving the problem of ensuring a generated subject remains the same even if it falls out of view temporarily.
OpenAI represents video as collections of smaller groups of data called “patches,” which the company says are similar to tokens (fragments of a word) in GPT-4. “By unifying how we represent data, we can train diffusion transformers on a wider range of visual data than was possible before, spanning different durations, resolutions, and aspect ratios,” the company writes.
An important tool in OpenAI’s bag of tricks is that its use of AI models is compounding. Earlier models are helping to create more complex ones. Sora follows prompts well because, like DALL-E 3, it utilizes synthetic captions that describe scenes in the training data generated by another AI model like GPT-4V. And the company is not stopping here. “Sora serves as a foundation for models that can understand and simulate the real world,” OpenAI writes, “a capability we believe will be an important milestone for achieving AGI.”
One question on many people’s minds is what data OpenAI used to train Sora. OpenAI has not revealed its dataset, but based on what people are seeing in the results, it’s possible OpenAI is using synthetic video data generated in a video game engine in addition to sources of real video (say, scraped from YouTube or licensed from stock video libraries). Nvidia’s Dr. Jim Fan, who is a specialist in training AI with synthetic data, wrote on X, “I won’t be surprised if Sora is trained on lots of synthetic data using Unreal Engine 5. It has to be!” Until confirmed by OpenAI, however, that’s just speculation.
Enlarge/ Still images of AI-generated video examples provided by Google for its Lumiere video synthesis model.
On Tuesday, Google announced Lumiere, an AI video generator that it calls “a space-time diffusion model for realistic video generation” in the accompanying preprint paper. But let’s not kid ourselves: It does a great job at creating videos of cute animals in ridiculous scenarios, such as using roller skates, driving a car, or playing a piano. Sure, it can do more, but it is perhaps the most advanced text-to-animal AI video generator yet demonstrated.
According to Google, Lumiere utilizes unique architecture to generate a video’s entire temporal duration in one go. Or, as the company put it, “We introduce a Space-Time U-Net architecture that generates the entire temporal duration of the video at once, through a single pass in the model. This is in contrast to existing video models which synthesize distant keyframes followed by temporal super-resolution—an approach that inherently makes global temporal consistency difficult to achieve.”
In layperson terms, Google’s tech is designed to handle both the space (where things are in the video) and time (how things move and change throughout the video) aspects simultaneously. So, instead of making a video by putting together many small parts or frames, it can create the entire video, from start to finish, in one smooth process.
The official promotional video accompanying the paper “Lumiere: A Space-Time Diffusion Model for Video Generation,” released by Google.
Lumiere can also do plenty of party tricks, which are laid out quite well with examples on Google’s demo page. For example, it can perform text-to-video generation (turning a written prompt into a video), convert still images into videos, generate videos in specific styles using a reference image, apply consistent video editing using text-based prompts, create cinemagraphs by animating specific regions of an image, and offer video inpainting capabilities (for example, it can change the type of dress a person is wearing).
In the Lumiere research paper, the Google researchers state that the AI model outputs five-second long 1024×1024 pixel videos, which they describe as “low-resolution.” Despite those limitations, the researchers performed a user study and claim that Lumiere’s outputs were preferred over existing AI video synthesis models.
As for training data, Google doesn’t say where it got the videos they fed into Lumiere, writing, “We train our T2V [text to video] model on a dataset containing 30M videos along with their text caption. [sic] The videos are 80 frames long at 16 fps (5 seconds). The base model is trained at 128×128.”
Enlarge/ A block diagram showing components of the Lumiere AI model, provided by Google.
AI-generated video is still in a primitive state, but it’s been progressing in quality over the past two years. In October 2022, we covered Google’s first publicly unveiled image synthesis model, Imagen Video. It could generate short 1280×768 video clips from a written prompt at 24 frames per second, but the results weren’t always coherent. Before that, Meta debuted its AI video generator, Make-A-Video. In June of last year, Runway’s Gen2 video synthesis model enabled the creation of two-second video clips from text prompts, fueling the creation of surrealistic parody commercials. And in November, we covered Stable Video Diffusion, which can generate short clips from still images.
AI companies often demonstrate video generators with cute animals because generating coherent, non-deformed humans is currently difficult—especially since we, as humans (you are human, right?), are adept at noticing any flaws in human bodies or how they move. Just look at AI-generated Will Smith eating spaghetti.
Judging by Google’s examples (and not having used it ourselves), Lumiere appears to surpass these other AI video generation models. But since Google tends to keep its AI research models close to its chest, we’re not sure when, if ever, the public may have a chance to try it for themselves.
As always, whenever we see text-to-video synthesis models getting more capable, we can’t help but think of the future implications for our Internet-connected society, which is centered around sharing media artifacts—and the general presumption that “realistic” video typically represents real objects in real situations captured by a camera. Future video synthesis tools more capable than Lumiere will make deceptive deepfakes trivially easy to create.
To that end, in the “Societal Impact” section of the Lumiere paper, the researchers write, “Our primary goal in this work is to enable novice users to generate visual content in an creative and flexible way. [sic] However, there is a risk of misuse for creating fake or harmful content with our technology, and we believe that it is crucial to develop and apply tools for detecting biases and malicious use cases in order to ensure a safe and fair use.”
On Monday, ChatGPT maker OpenAI detailed its plans to prevent the misuse of its AI technologies during the upcoming elections in 2024, promising transparency in AI-generated content and enhancing access to reliable voting information. The AI developer says it is working on an approach that involves policy enforcement, collaboration with partners, and the development of new tools aimed at classifying AI-generated media.
“As we prepare for elections in 2024 across the world’s largest democracies, our approach is to continue our platform safety work by elevating accurate voting information, enforcing measured policies, and improving transparency,” writes OpenAI in its blog post. “Protecting the integrity of elections requires collaboration from every corner of the democratic process, and we want to make sure our technology is not used in a way that could undermine this process.”
Initiatives proposed by OpenAI include preventing abuse by means such as deepfakes or bots imitating candidates, refining usage policies, and launching a reporting system for the public to flag potential abuses. For example, OpenAI’s image generation tool, DALL-E 3, includes built-in filters that reject requests to create images of real people, including politicians. “For years, we’ve been iterating on tools to improve factual accuracy, reduce bias, and decline certain requests,” the company stated.
OpenAI says it regularly updates its Usage Policies for ChatGPT and its API products to prevent misuse, especially in the context of elections. The organization has implemented restrictions on using its technologies for political campaigning and lobbying until it better understands the potential for personalized persuasion. Also, OpenAI prohibits creating chatbots that impersonate real individuals or institutions and disallows the development of applications that could deter people from “participation in democratic processes.” Users can report GPTs that may violate the rules.
OpenAI claims to be proactively engaged in detailed strategies to safeguard its technologies against misuse. According to their statements, this includes red-teaming new systems to anticipate challenges, engaging with users and partners for feedback, and implementing robust safety mitigations. OpenAI asserts that these efforts are integral to its mission of continually refining AI tools for improved accuracy, reduced biases, and responsible handling of sensitive requests
Regarding transparency, OpenAI says it is advancing its efforts in classifying image provenance. The company plans to embed digital credentials, using cryptographic techniques, into images produced by DALL-E 3 as part of its adoption of standards by the Coalition for Content Provenance and Authenticity. Additionally, OpenAI says it is testing a tool designed to identify DALL-E-generated images.
In an effort to connect users with authoritative information, particularly concerning voting procedures, OpenAI says it has partnered with the National Association of Secretaries of State (NASS) in the United States. ChatGPT will direct users to CanIVote.org for verified US voting information.
“We want to make sure that our AI systems are built, deployed, and used safely,” writes OpenAI. “Like any new technology, these tools come with benefits and challenges. They are also unprecedented, and we will keep evolving our approach as we learn more about how our tools are used.”
Enlarge/ An AI-generated image of a “Beautiful queen of the universe looking at the camera in sci-fi armor, snow and particles flowing, fire in the background” created using alpha Midjourney v6.
Midjourney
In December, just before Christmas, Midjourney launched an alpha version of its latest image synthesis model, Midjourney v6. Over winter break, Midjourney fans put the new AI model through its paces, with the results shared on social media. So far, fans have noted much more detail than v5.2 (the current default) and a different approach to prompting. Version 6 can also handle generating text in a rudimentary way, but it’s far from perfect.
“It’s definitely a crazy update, both in good and less good ways,” artist Julie Wieland, who frequently shares her Midjourney creations online, told Ars. “The details and scenery are INSANE, the downside (for now) are that the generations are very high contrast and overly saturated (imo). Plus you need to kind of re-adapt and rethink your prompts, working with new structures and now less is kind of more in terms of prompting.”
At the same time, critics of the service still bristle about Midjourney training its models using human-made artwork scraped from the web and obtained without permission—a controversial practice common among AI model trainers we have coveredindetail in the past. We’ve also covered the challenges artists might face in the future from these technologies elsewhere.
Too much detail?
With AI-generated detail ramping up dramatically between major Midjourney versions, one could wonder if there is ever such as thing as “too much detail” in an AI-generated image. Midjourney v6 seems to be testing that very question, creating many images that sometimes seem more detailed than reality in an unrealistic way, although that can be modified with careful prompting.
An AI-generated image of a nurse in the 1960s created using alpha Midjourney v6.
Midjourney
An AI-generated image of an astronaut created using alpha Midjourney v6.
Midjourney
An AI-generated image of a “juicy flaming cheeseburger” created using alpha Midjourney v6.
Midjourney
An AI-generated image of “a handsome Asian man” created using alpha Midjourney v6.
Midjourney
An AI-generated image of an “Apple II” sitting on a desk in the 1980s created using alpha Midjourney v6.
Midjourney
An AI-generated image of a “photo of a cat in a car holding a can of beer” created using alpha Midjourney v6.
Midjourney
An AI-generated image of a forest path created using alpha Midjourney v6.
Midjourney
An AI-generated image of a woman among flowers created using alpha Midjourney v6.
Midjourney
An AI-generated image of “a plate of delicious pickles” created using alpha Midjourney v6.
Midjourney
An AI-generated image of a barbarian beside a TV set that says “Ars Technica” on it created using alpha Midjourney v6.
Midjourney
An AI-generated image of “Abraham Lincoln holding a sign that says Ars Technica” created using alpha Midjourney v6.
Midjourney
An AI-generated image of Mickey Mouse holding a machine gun created using alpha Midjourney v6.
Midjourney
In our testing of version 6 (which can currently be invoked with the “–v 6.0” argument at the end of a prompt), we noticed times when the new model appeared to produce worse results than v5.2, but Midjourney veterans like Wieland tell Ars that those differences are largely due to the different way that v6.0 interprets prompts. That is something Midjourney is continuously updating over time. “Old prompts sometimes work a bit better than the day they released it,” Wieland told us.
“Here, There, and Everywhere” isn’t just a Beatles song. It’s also a phrase that recalls the spread of generative AI into the tech industry during 2023. Whether you think AI is just a fad or the dawn of a new tech revolution, it’s been impossible to deny that AI news has dominated the tech space for the past year.
We’ve seen a large cast of AI-related characters emerge that includes tech CEOs, machine learning researchers, and AI ethicists—as well as charlatans and doomsayers. From public feedback on the subject of AI, we’ve heard that it’s been difficult for non-technical people to know who to believe, what AI products (if any) to use, and whether we should fear for our lives or our jobs.
Meanwhile, in keeping with a much-lamented trend of 2022, machine learning research has not slowed down over the past year. On X, former Biden administration tech advisor Suresh Venkatasubramanianwrote, “How do people manage to keep track of ML papers? This is not a request for support in my current state of bewilderment—I’m genuinely asking what strategies seem to work to read (or “read”) what appear to be 100s of papers per day.”
To wrap up the year with a tidy bow, here’s a look back at the 10 biggest AI news stories of 2023. It was very hard to choose only 10 (in fact, we originally only intended to do seven), but since we’re not ChatGPT generating reams of text without limit, we have to stop somewhere.
Bing Chat “loses its mind”
Aurich Lawson | Getty Images
In February, Microsoft unveiled Bing Chat, a chatbot built into its languishing Bing search engine website. Microsoft created the chatbot using a more raw form of OpenAI’s GPT-4 language model but didn’t tell everyone it was GPT-4 at first. Since Microsoft used a less conditioned version of GPT-4 than the one that would be released in March, the launch was rough. The chatbot assumed a temperamental personality that could easily turn on users and attack them, tell people it was in love with them, seemingly worry about its fate, and lose its cool when confronted with an article we wrote about revealing its system prompt.
Aside from the relatively raw nature of the AI model Microsoft was using, at fault was a system where very long conversations would push the conditioning system prompt outside of its context window (like a form of short-term memory), allowing all hell to break loose through jailbreaks that people documented on Reddit. At one point, Bing Chat called me “the culprit and the enemy” for revealing some of its weaknesses. Some people thought Bing Chat was sentient, despite AI experts’ assurances to the contrary. It was a disaster in the press, but Microsoft didn’t flinch, and it ultimately reigned in some of Bing Chat’s wild proclivities and opened the bot widely to the public. Today, Bing Chat is now known as Microsoft Copilot, and it’s baked into Windows.
US Copyright Office says no to AI copyright authors
Enlarge/ An AI-generated image that won a prize at the Colorado State Fair in 2022, later denied US copyright registration.
Jason M. Allen
In February, the US Copyright Office issued a key ruling on AI-generated art, revoking the copyright previously granted to the AI-assisted comic book “Zarya of the Dawn” in September 2022. The decision, influenced by the revelation that the images were created using the AI-powered Midjourney image generator, stated that only the text and arrangement of images and text by Kashtanova were eligible for copyright protection. It was the first hint that AI-generated imagery without human-authored elements could not be copyrighted in the United States.
This stance was further cemented in August when a US federal judge ruled that art created solely by AI cannot be copyrighted. In September, the US Copyright Office rejected the registration for an AI-generated image that won a Colorado State Fair art contest in 2022. As it stands now, it appears that purely AI-generated art (without substantial human authorship) is in the public domain in the United States. This stance could be further clarified or changed in the future by judicial rulings or legislation.
