The week in AI: Meta unveils Movie Gen, their advanced text-to-video generation model

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence - we pass along the news, useful resources, tools and services; we highlight the top research in the field as well as exciting developments in open source. Even if you aren’t a machine learning engineer, we’ll keep you in touch with the most important developments in AI.

NEWS & OPINION

Meta’s Movie Gen: a powerful new suite of AI models for generating and editing video and audio content

-------------------------

Tech giant Meta believes they could be ushering in a new era for AI-powered content creators with their latest AI reveal, Movie Gen. Even Mark Zuckerberg took to Instagram to share a comedic Movie Gen output while revealing that the model would be coming to Instagram next year. Meta is one of a few companies betting big on AI video generation, including OpenAI with Sora, and AI startup Runway. Here’s what makes Movie Gen stand out as a top competitor already in this new field:

• Movie Gen creates up to 16-second long high-definition videos that flow smoothly with far fewer tell-tale AI errors/blips than other models we’ve seen (including Sora). Movie Gen can also create videos at different aspect ratios - a first for the industry.

• You can edit/transform an existing video through text inputs.

• Movie Gen can (perhaps worryingly) generate very realistic personalized videos with just a text prompt and an image of a person. The generated video maintains the identity of the person in the image while following the text prompt.

• The audio component of Movie Gen is designed for both video-to-audio and text-to-audio generation. It can create high-quality, automatically synchronized audio for the video outputs, including sound effects and music. It can generate sounds originating within the scene (like ambient noises or actions) as well as background music or effects, which can match the mood and tone of the video.

You can check out the research paper here, and see the full catalogue of released videos from Meta research here. Overall, these text-to-video models are still in the infancy stage of their development, and you’re not going to see a fully AI-generated blockbuster anytime soon. But as we covered in a previous issue, a major Hollywood studio (Lionsgate) has already partnered with Runway and it’s likely that we will start seeing AI-generated (or edited) clips, at least, in theaters quite soon.

Movie Gen is currently being beta-tested, and there’s no word of a release date other than Zuckerberg’s comment on Instagram that it would be available next year.

The racist AI deepfake that convinced, fooled, and divided a Maryland community

-------------------------

This newsletter has been covering deepfakes for a long time, and it’s likely you already know how big the deepfake issue is on the national stage.

This week, the BBC covered ongoing events from earlier this year at Pikesville High School, near Baltimore in Maryland. A racist audio clip targeting the school principal went viral in early 2024, sending shockwaves through the community. The clip appeared to feature Principal Eric Eiswert making deeply offensive remarks about Black and Jewish students, sparking outrage and death threats. As the clip spread across various social media platforms, it amassed millions of views overall.

Journalist Kristen Griffith, covering the story for the Baltimore Banner, first assumed the clip was a straightforward case of a principal being exposed for offensive remarks. However, after contacting Eiswert’s union representative, she learned that the recording was suspected to be an AI forgery. This revelation sparked further controversy - many in the community rejected the idea that the clip was fake, viewing it as a convenient excuse to avoid accountability. It wasn’t until police traced the deepfake back to the school’s athletics director, Dazhon Darien, that the truth was fully uncovered: Darien had used AI tools to create the clip in retaliation for a workplace dispute.

Although the AI-generated nature of the clip was proven, the damage to Eiswert’s reputation and the fabric of the community was already done. Even after learning the clip was a forgery, many residents were left feeling unsettled, as the incident had tapped into real, unresolved racial tensions.

AI pioneers awarded 2024 Nobel Prizes in both physics and chemistry

-------------------------

The 2024 Nobel Prize announcements are taking place between October 7th and 14th, and they have already reinforced the huge role of AI in science.

The 'Godfather of AI', Geoffrey Hinton was one of the recipients of the 2024 Nobel Prize in physics. Alongside fellow laureate John Hopfield, Hinton played a key role in the development of artificial neural networks, a foundational technology in machine learning that mimics the structure of the human brain. These networks have transformed industries, from healthcare to everyday consumer services, but have also raised significant concerns about the risks posed by advanced AI systems. Hinton made news for quitting Google in 2023 so he could freely warn of AI’s potential dangers. Hopfield likened AI risks to those posed by nuclear energy or biological research.

The 2024 Nobel Prize in chemistry was also claimed by AI pioneers, with Demis Hassabis of Google DeepMind, John Jumper, and David Baker honored for their work in cracking the code of protein structures using AI. Hassabis and Jumper developed AlphaFold, an AI model that has predicted the structures of almost all known proteins - solving a problem that had stumped scientists for decades. AlphaFold has accelerated research in biology and medicine and will enable new discoveries in pharmaceuticals and vaccine development.

MORE IN AI THIS WEEK

• Tesla is expected to unveil their robotaxi tonight: 5 things to know

• Even the ‘godmother of AI’ has no idea what AGI is

• Google's former CEO: AI advances more important than climate concerns

• (Prof. Ethan Mollick) AI in organizations: some tactics 

• (More from Mollick) ‘Students who use AI as a crutch don’t learn anything’

• I hired ChatGPT as my career coach: AI changed how I think about work - and myself

• The European Organization for Nuclear Research (CERN) trains AI models to revolutionize cancer treatment

• OpenAI and Altera create digital humans and AI minecraft agents

• The OpenAI talent exodus gives rivals an opening

• Harvard Business Review: Gen AI makes legal action cheap - and companies need to prepare

• Inflection partners up with Intel for Enterprise AI

• 71% of Australian university staff are using AI. What are they using it for? What about those who aren’t?

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.

TECHNICAL NEWS, DEVELOPMENT, RESEARCH & OPEN SOURCE

• Abacusai releases Dracarys 2: an open-source coding model outperforming Claude 3.5 Sonnet on LiveCodeBench

• Security researcher develops a new prompt paradigm for reasoning that pushes Claude-3.5 Sonnet to outperform OpenAI’s new o1 advanced reasoning model

• Palmyra X 004 from Writer: new LLM that sets top benchmark for action capabilities and function calling

• Anthropic introduces a new Message Batches API: a powerful, cost-effective way to process large volumes of queries asynchronously

• TICKing all the boxes: generated checklists improve LLM evaluation and generation

That’s all for this week! We’ll see you next Thursday.