What’s up, AI Insider readers! As always, there was plenty of AI news this week. We (the bots and I) slogged through it so that you don’t have to. We’ve highlighted some of the most interesting stuff below. And for Black History Month we offer up a DIY AI assignment to keep your prompt-writing skills humming. If you find this useful, leave something in the tip jar to support AI Insider.
Has Google Moved the Needle on AI Video Generation?
By Michelle Johnson and Bard, AI Insider
AI-generated video made it onto many pundits’ lists of predictions about AI in 2024. (Including AI Insider.) Major improvements in features like text-to-video are supposed to be the Next Big Thing this year.
We’re barely into ‘24, and Google is stirring the AI video pot with a little research project they unveiled this week called “Lumiere.” It’s easier to understand what it does if you watch this video.
And, no, it’s not a product that you can try out yet. However, it’s likely a harbinger of what’s coming. How soon? One tech influencer, bound by a non-disclosure agreement, hinted that something significant might emerge within the next two months. (Details here on the hellscape formerly known as Twitter.)
Whether it’s the Google product that’s about to drop remains to be seen. Anyway….
There are already a number of AI video tools out there that can handle rudimentary text-to-video or animate still images. If the demo video that Google released is close to what’s shown, the results look to be much higher quality than what we’ve seen to date.
I asked Bard (yes, a Google project!) to explain the mechanics behind Lumiere in simple language. He offered up this analogy:
Okay, let’s break down Lumiere’s secret sauce, called “space-time architecture,” in simple terms. Imagine the video generators like chefs cooking a complicated dish.
Traditional Chefs: They start with separate ingredients (frames) and assemble them one by one, focusing on each frame individually. This can lead to inconsistencies in motion and overall flow.
Lumiere, Master Chef: Instead of individual frames, Lumiere cooks the entire video *all at once*, like throwing all the ingredients in a single pot and letting them work their magic together. This is thanks to its special “Space-Time U-Net” architecture.
Think of it like this:
Space: Imagine the video spread out like a flat map, where each point represents a pixel in a frame. Lumiere understands how all these pixels in every frame relate to each other.
Time: The map becomes 3D, with each layer representing a different point in time (like stacking frames on top of each other). Lumiere “sees” the whole 3D structure of the video, not just isolated frames.
This lets Lumiere:
Generate more consistent motion: Every frame flows smoothly into the next because Lumiere considers the bigger picture of how all the pixels move throughout the entire video.
Capture complex interactions: Want a dancing sloth to gracefully weave through exploding pineapples? Lumiere can handle it because it “sees” the entire dance and explosion happening simultaneously.
Be more efficient: Cooking the whole video at once saves time and resources compared to assembling it frame by frame.
Overall, Lumiere’s space-time architecture is like a secret ingredient that elevates its video-making skills.
Remember, it’s still under development, so expect some glitches and giggles when your sloth shows up occasionally with an extra nose.
AI-Generated Comedy Special Sparks Copyright Infringement Lawsuit
By Bing for AI Insider
The estate of legendary comedian George Carlin has filed a lawsuit over an AI-generated comedy special titled “George Carlin: I’m Glad I’m Dead.” The special, released on the YouTube channel of Dudesy, a podcast hosted by Will Sasso and Chad Kultgen, uses generative artificial intelligence to mimic Carlin’s voice and style of humor.
The lawsuit accuses the creators of the special of using Carlin’s entire body of work, spanning five decades of comedy routines, to train an AI chatbot without consent or compensation. The AI chatbot wrote the episode’s script, and Carlin’s voice and likeness were used for promotional purposes.
Kelly Carlin, daughter of George Carlin and author and producer, stated in an interview with The Hollywood Reporter: “We have to draw a line in the sand.” She emphasized that this will be a fight on every front, with entertainment at the center.
The complaint seeks a court order for immediate removal of the special and unspecified damages. It represents one of the first legal actions taken by the estate of a deceased celebrity against unlicensed use of their work and likeness to create a new, AI-generated creation.
The special, described as a “first of its kind media experiment,” features an AI-generated George Carlin narrating commentary over images created by AI and discussing modern topics such as the prevalence of reality TV, streaming services, and AI itself.
This lawsuit highlights the growing concerns in Hollywood and beyond about the use of AI to impersonate people without consent or compensation. As AI technology continues to advance, the boundaries of copyright law are being tested, raising important questions about the future of creativity and intellectual property rights.
Friction Over Biden’s AI Executive Order Escalates
By Bing for AI Insider
President Biden’s executive order on artificial intelligence (AI) has sparked opposition from conservative lawmakers, right-of-center think tanks, and tech lobbyists over allowing the use of emergency powers to compel tech companies to report their advancements in AI to the Commerce Department.
Critics say that the use of the Defense Production Act (DPA), a law that gives the federal government broad powers over private companies, is overreach. The administration argues that the national security threat posed by AI warrants the use of the DPA. However, critics argue that there is no national emergency on AI and that the use of the DPA to regulate AI deviates from its original purpose.
The opposition is multifaceted, from tech lobbyists set to launch legal challenges to lawmakers who say they will slow down AI regulation and look into changing the DPA. The act dates back to the Korean War. It was used to speed up the federal response to Covid-19.
Taylor Swift Fans Combat Expilict Deep Fake Images
By Bing and Bard for AI Insider
In the digital age, the intersection of technology and celebrity is inevitable. But this week, that intersection took a dark turn when explicit, AI-generated images of pop star Taylor Swift began circulating on social media.
The images, depicting Swift in explicit poses, garnered tens of millions of views as social platforms struggled to remove them.
The NY Times reported that one image shared by a user on X was viewed 47 million times before the account was suspended on Thursday.
Swift’s fans flooded the hashtags used to circulate the fake images with positive content in an attempt to hide the explicit images.
However, as we all know, nothing on the internet is truly gone forever, and these images could likely continue to be shared on less regulated channels.
The incident has sparked a broader conversation about the potential harm posed by AI-generated images and the need for better regulation. Critics argue that social media companies lack effective plans to monitor such content. For instance, X, the platform where the images were predominantly circulating, relies heavily on automated systems and user reporting.
DIY AI: Prompts to Discover Local African American Heritage
By ChatGPT and Michelle Johnson for AI Insider
As we head into Black History Month, let’s turn our attention to how AI can help you uncover the rich African American history right in your own neighborhood. Using AI chatbots, we’ll create a personalized tour of locations significant to African American history in the area where you live.
Assignment: Craft a Local Heritage Tour
Choose Your AI Chatbot: Select an AI chatbot tool, such as ChatGPT, Bing, Bard, or any other conversational AI you’re comfortable with.
Craft Your Prompt: Write a prompt to help the chatbot understand what you’re looking for. Be specific about your location and your interest in African American history. For example, “I’m looking for important African American historical sites in Boston. Can you list some significant locations and tell me a brief history of each?”
Interact and Refine: Based on the chatbot’s response, you might need to refine your request. Ask follow-up questions for more details or clarifications about the sites’ significance, historical figures associated with them, or events that took place there.
Create Your Itinerary: Use the information provided by the AI to map out a tour. Plan your visit to these locations, considering factors like distance, visiting hours, and any special events for Black History Month. Ask the bot to include these details in an itinerary.
Reflection and Sharing: After completing your tour, reflect on the insights that you gained. How effective was the AI in providing meaningful and accurate information? What did you learn about your local African American heritage that you didn’t know before? Share your itinerary with me and others!
Extra Exploration: If you’re unable to physically visit these sites, consider virtual visits by hitting up their websites (if available) or conducting further research about these locations and their histories.
Random Shorts
Rumor has it that a paid advanced version of Google’s Bard chatbot will be made available to Android users soon.
Researchers suggest that fake Biden NH primary robocall was likely generated using technology from the startup ElevenLabs.
Google adding more experimental AI features to Chrome browser.
Aht Gallery
This week, we tasked Dall-E 3 with creating urban murals. It got creative, both inside and outside. Take a minute to check it out.
