Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 👏 OpenAI claps back against Elon Musk

• 📜 Claude 3.5 Haiku released

• 🎥 Meta debuts a tool for watermarking AI-generated videos

• 📚 Harvard releases 1 million public-domain books for AI training

Let’s get right into it!

Sign Up | Advertise | Tools

STARTUPS

OpenAI claps back against Elon Musk

OpenAI has responded to Elon Musk's legal efforts to halt its transition to a fully for-profit model, alleging that Musk himself advocated for a for-profit structure in 2017 but walked away when he couldn’t gain majority control. Musk’s xAI, a competitor to OpenAI, has since raised significant funds and launched its Grok chatbot.

OpenAI, which started as a nonprofit in 2015 and transitioned to a "capped-profit" model in 2019, has seen massive growth, largely fueled by ChatGPT, raising billions from investors like Microsoft. Musk has accused OpenAI of breaching fiduciary duties and engaging in anti-competitive practices, claiming its ties with Microsoft disadvantage competitors like xAI.

OpenAI denies these allegations, stating its investors are free to invest elsewhere, though they may lose access to proprietary information. The legal feud underscores the fierce competition and high stakes in the booming generative AI market, projected to surpass $1 trillion in revenue within a decade.

STARTUPS

Claude 3.5 Haiku

Anthropic has made its latest AI model, Claude 3.5 Haiku, generally available to all chatbot users. Announced in October, Haiku is the company's fastest model, optimized for tasks like coding, data extraction, content moderation, personalization, and scalable chatbot development.

While benchmarks place Haiku below Claude 3 Opus and 3.5 Sonnet, it is reportedly more cost-effective, faster, and higher quality than many competitors, especially for coding applications. It also integrates with Artifacts, a feature that displays real-time, formatted results alongside conversations.

BIG TECH

Meta debuts a tool for watermarking AI-generated videos

The rise of generative AI has led to a surge in deepfake content, with deepfakes accounting for 7% of global fraud in 2024, according to Sumsub. To combat this, Meta has launched Video Seal, an open-source tool for applying imperceptible watermarks to AI-generated videos, making it easier to detect and trace their origins.

Unlike other solutions, Video Seal is designed to withstand compression, blurring, and cropping while offering scalability and efficiency. It also supports hidden messages for verifying content authenticity. Despite its promise, adoption challenges remain, as many developers already use proprietary tools. To encourage widespread use, Meta is launching a public leaderboard and hosting a workshop at ICLR to advance collaboration in watermarking technology.

EDUCATION

Harvard releases 1 million public-domain books for AI training

Harvard University has unveiled a high-quality dataset of nearly 1 million public-domain books, created by its Institutional Data Initiative with funding from Microsoft and OpenAI. The dataset, sourced from Google Books, includes diverse works ranging from Shakespeare to niche textbooks and dictionaries, offering a valuable resource for training AI models.

The initiative aims to "level the playing field" by providing smaller players in the AI industry and individual researchers access to curated content, similar to what tech giants use. While proponents highlight its potential to reduce reliance on copyrighted materials, critics warn that its effectiveness depends on replacing, rather than supplementing, scraped data.