This post was featured in the Reality Defender Newsletter. To receive news, updates, and more on deepfakes and Generative AI in your inbox, subscribe to the Reality Defender Newsletter today.
Deepfake of the Week
Canadian artists Drake and The Weeknd have a fairly extensive history of collaborating with each other on several songs. So when the song “Heart On My Sleeve” appeared on streaming services from the mysterious artist Ghostwriter, fans were excited for all of five seconds that the two blockbuster artists got together for the first time in a decade. Until, of course, news quickly broke that the song synthesized both artists’ voices and was (unfortunately) a fake.
What does Reality Defender's platform say?
AI-created voices are easy for our platform to detect, though the underlying music made this one a tad trickier than your run-of-the-mill isolated voice deepfakes. Fortunately, our models are constantly trained and retrained to factor in everything from several layers of transcoding to a “Scorpion-era Drake Type Beat.”
What does Reality Defender's team say?
Get ready for the onslaught of AI-generated music, using simulated voices of the artists you know and love. Universal is already proactively asking streaming platforms to avoid letting models train on their music, which is great in theory but doesn’t factor in smaller, open-source models training on illegally downloaded songs.
Not all AI-generated music is made with an intent to deceive and mimic. (For instance, Holly Herndon’s Proto is well-written electronic music and AI generated.) Yet if models can make a convincing-enough Drake and Weeknd collaboration that goes viral, we’re not far off from an onslaught of deepfaked songs flooding streaming platforms.
As demonstrated above, Reality Defender’s platform is equipped to detect these tracks to help artists, managers, and rights holders stop advanced theft of their intellectual property.
Open (Source) AI
Want to roll your own LLM? Now you can with Dolly, an open-source alternative to ChatGPT. Unlike ChatGPT, Dolly requires training on a user’s own machine and/or instance, while also allowing users to potentially circumvent the content moderation and abuse checks found in ChatGPT, Bardo, Bing AI, and so on. Dolly’s launch comes after Meta’s own partially-open sourced LLM, and alongside AutoGPT, an open-sourced attempt to give users a way to automate functions of GPT4.
Generative AI Comes to AWS
Amazon announced Amazon Bedrock and Titan, a new way to “build and scale generative AI applications with FMs.” Upon Bedrock’s introduction, CEO Andy Jassy spoke to CNBC about the billion-dollar cost of training “really good” AI models.
Poll of the Week
Which type of deepfake has the greatest potential to disrupt and/or damage your business when weaponized by bad actors?
Elon Musk Registers AI Company
X.AI is Elon Musk’s attempt at starting an AI company. Musk’s registration of the company comes after news that the Twitter and Tesla CEO bought thousands of high-powered GPUs. X.AI’s newfound existence also coincides with the recent Musk-led open letter to pause development on “giant AI experiments.”
More News
- The U.S. is seeking to quickly establish laws concerning the use and proliferation of AI. (The Guardian)
- A new report from industry researchers shows the direct link between the rise of AI and the “concentration of power” from big tech companies. (AI Now Institute)
- The use of deepfakes from state-run media is continuing to rise. (The Conversation)
- Spending on AI-powered apps is now up 4000%. (TechCrunch)
- Meta’s stance on political ads utilizing deepfakes remains unclear. (The Washington Post)
- China is allegedly taking a cautious, review-first approach to AI service before they’re released to the public. (Quartz)
- Meanwhile, a China-based media and public relations group is now replacing humans with LLMs. (Bloomberg)
- An updated version of Stable Diffusion is now available to use. (Fast Company)
- Your next piece of furniture from Ikea could be generated by AI. (Yanko Design)
- AI-generated porn could soon be illegal in California. (Fox News)
- Sundar Pichai of Alphabet is calling for a global framework to combat harmful deployment of AI. (The Guardian)
Thank you for reading the Reality Defender Newsletter. If you have any questions about Reality Defender, or if you would like to see anything in future issues, please reach out to us here.
