Good morning. It’s Monday, October 21st.

Did you know: On this day in 1991, Apple Computer introduced the first PowerBook laptops for $2,500 (equivalent to $5,700 in 2024)

In today’s email:

• OpenAI’s o1 Model Talk

• Mira seeking $100M For AI Startup

• Meta's Spirit LM

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

_{In partnership with: Lead Magnet Generator}

Generate High-Converting Lead Magnets in Minutes!

For small business owners who need more leads without the hassle: The Lead Magnet Generator creates professional ebooks and whitepapers, automatically—no design or writing skills needed.

Here’s What You’ll Get: - Content That Converts: Just a few clicks, and you have a polished, ready-to-use lead magnet in PDF format. - Limited-Time Launch Price: Lock in a low rate from just $14 before prices go up permanently.

Available for 3 days only. 14-Day Satisfaction Guarantee!

Not seeing results? Get 2 new lead magnets for free.

Grab your spot and start getting more leads!

_{Thank you for supporting our sponsors!}

OpenAI’s o1 Model May Make Most Lawyers Obsolete

At Ray Summit 2024, OpenAI’s Chief Product Officer, Kevin Weil, discussed the o1 model's ability to draft legal briefs in minutes, a task that previously took six hours of a $1,000-per-hour associate’s time. This efficiency represents a 99.9% reduction in both time and cost, prompting a reassessment of traditional pricing models in professional services.

Weil’s Ray Summit 2024 full video [36min]:

Weil underscored OpenAI's mission to democratize AI access while juggling the need for premium services for those who can afford them. This balancing act raises essential questions about aligning value creation with sustainable monetization, all while ensuring global accessibility. Read more.

Ex-OpenAI CTO Mira Murati reportedly seeking over $100M to develop proprietary models

Former OpenAI CTO Mira Murati is reportedly on the hunt for over $100 million to launch her own AI startup, joining a growing list of ex-OpenAI executives venturing into the wild frontier of independent projects.

Discussions with venture capitalists are underway to develop AI products based on proprietary models, though it remains unclear whether Murati will assume the CEO position. Her experience at OpenAI, where she contributed to key projects like ChatGPT and DALL-E, positions her favorably in this competitive market.

Currently in discussions with venture capitalists, Murati aims to develop AI products based on proprietary models, although it’s unclear if she’ll take the helm as CEO. Her experience at OpenAI, where she played a pivotal role in projects like ChatGPT and DALL-E, gives her a solid edge in a saturated market.

Meanwhile, other notable alumni, such as Ilya Sutskever and Andrej Karpathy, are also pursuing new endeavors, prompting questions about OpenAI’s market stronghold. As these former insiders take their expertise elsewhere, it raises eyebrows regarding whether the company truly holds a monopoly on AI innovation or if there are cracks in the facade. Read more.

Meta's Spirit LM generates more expressive voices that reflect anger, surprise, happiness and other emotions

Meta's Fundamental AI Research team has launched Spirit LM, an open-source multimodal language model that steps up the game in AI voice technology. Unlike traditional models that often resemble robots reading a script, Spirit LM uses phonetic, pitch, and tone tokens to add a splash of emotional flair to its outputs.

The model comes in two varieties: Spirit LM Base, which nails the basics of phonetic processing, and the more sophisticated Spirit LM Expressive, capable of embodying feelings like joy, anger, and surprise. Trained on diverse text and speech datasets, Spirit LM offers humanlike text-to-speech and speech-to-text capabilities.

The models were trained on diverse datasets and are released under the FAIR Noncommercial Research License, enabling the research community to explore their potential. Additionally, Meta has updated its Segment Anything model for enhanced image and video segmentation tasks. Read more or request early access here.

• DeepMind researchers find LLMs can serve as effective mediators

• New Research by Anthropic and Stanford Investigates LLM Introspection and Its Implications for AI Ethics

• Janus AI model fuses image understanding and generation in a single adaptable framework

• Microsoft releases framework for highly efficient 1-bit language models

• OpenAI, Microsoft reportedly hire banks to renegotiate partnership terms 

• Nvidia, Google, Microsoft and more head to Las Vegas to tout health-care AI tools  

• Perplexity is reportedly looking to fundraise at an $8B valuation

• IBM debuts open source Granite 3.0 LLMs for enterprise AI

• In-context learning proves competitive with LLM fine-tuning when data is scarce

📄 Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation

📄 SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs  ​​

📄 L3DG: Latent 3D Gaussian Diffusion

📄 The Curse of Multi-Modalities: Evaluating Hallucinations of Large Multimodal Models across Language, Visual, and Audio

📄 Roadmap towards Superhuman Speech Understanding using Large Language Models

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!