. . AI: Very cool AI research tool, Mistral CEO on regulation, media generation, and more (11.16.23)

Friends, some really interesting things being discussed by the AI community today. I've been hearing that people want a little more context with their news summaries, so I've added a line here and there of my thoughts and related resources.

These are almost in least-interesting to most-interesting order for me today, make sure to check out the last story on LangChain's new research assistant template!

If you know someone who would appreciate one or all of these stories, I’d be very thankful if you forwarded it to them.

Now here’s today’s news,

Marshall Kirkpatrick, Editor

DeepMind and YouTube Launch Advanced AI Music Model, Lyria

First impacted: Music creators, Music producers
Time to impact: Medium

Google DeepMind and YouTube have launched Lyria, an AI model that can create high-quality music, and two AI experiments: Dream Track and Music AI tools. Dream Track lets creators make soundtracks with AI-generated voice and style, while Music AI tools help artists, songwriters, and producers transform audio from one style or instrument to another. The demos are all very short right now, but I can imagine a future where high quality music creation is so easy and abundant that music's ability to deepen or change our emotional experiences will be taken to a whole new level. The social and kinesthetic costs seem daunting, though. [Transforming the future of music creation] Share by email

Meta AI Launches Generative Tools Emu Video and Emu Edit

First impacted: Video editors, Social media content creators
Time to impact: Medium

Meta AI has introduced Emu Video and Emu Edit, two AI tools for video generation and image editing respectively. In a survey, 96% preferred Emu Video over Meta's previous Make-A-Video for quality and 85% for text prompt accuracy. The company says "the potential use cases are clearly evident:" clever GIFs and animated Instagram photos. So too, probably, did Guttenberg fail to foresee the Protestant Reformation. [insert snarky GIF] [Introducing Emu Video and Emu Edit, our latest generative AI research milestones] Share by email

Mistral AI CEO Criticizes EU AI Act's Shift in Focus

First impacted: Policymakers & open source AI community
Time to impact: Medium

Arthur Mensch, CEO of Mistral AI, has written an extended and well articulated post on X criticizing the EU AI Act's new focus on regulating "founditional models" and "systemic risks", arguing it should prioritize product safety and that its vague decision criteria may harm smaller firms. For further exploration of this perspective, see also Joanna Bryson's just rereleased conversation with Azeem Azhar on the technical side of AI safety.  [Via @arthurmensch] Share by email

Codegen Raises $16.2M to Automate Software Engineering Tasks

First impacted: Developers
Time to impact: Short to medium

Ex-Palantir AI researcher Jay Hack's startup Codegen has raised $16.2 million in "seed" funding. “Codegen leverages a multi-agent system for complex code generation,” Hack told TechCrunch. “This entails orchestrating a swarm of agents that collaboratively decompose and solve large tasks. Many LLMs effectively deliberate and build upon each other’s work, [which] yields significantly better outputs.” The startup's 1 minute demo video on X is something better seen than described.  [Codegen raises new cash to automate software engineering tasks | TechCrunch] Share by email

LangChain Launches Research Assistant Template

First impacted: AI savvy researchers
Time to impact: Short to medium

Moving beyond realtime chat, LangChain has introduced a new "Research Assistant" template, a tool designed to help with extensive research tasks. Using OpenAI for language modeling and Tavily for searching, it takes a user's query, generates sub-questions, researches and summarizes relevant documents, and compiles these summaries into a final (1st draft) report. If cognitive labor is what takes us from data, to information, to knowledge, to wisdom, then this feels like an AI offering to help with knowledge creation through topic decomposition, contextualization, and synthesis. Exciting! You can see a 4 minute video demo of part of the technology here.  ["Research Assistant": Exploring UXs Besides Chat] Share by email

That’s it! More AI news tomorrow!