Good morning. It’s Friday, November 8th.

Did you know: On this day in 2006, Windows Vista was released?

In today’s email:

• Google’s Jarvis Leak

• Kling’s Video Character Consistency Training

• Breakdown of all the AI news in the last 48hrs

• 4 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 Bagel}

🔥 Unlock the Future of AI Reasoning with Bagel! 🔥

AI is transforming our world like the discovery of fire, and Bagel is at the forefront of this revolution.

In Bagel’s latest research, they explore advancing Large Language Models from mere pattern recognizers to true cognitive agents capable of human-like reasoning. Explore how Bagel is enhancing arithmetic, commonsense, and symbolic reasoning through groundbreaking training and inference-time techniques.

Curious about the next leap in AI evolution?

Check out Bagel’s full blog post and join the journey towards smarter, more intuitive AI!

Google accidentally leaked a preview of its Jarvis AI that can take over computers

Google inadvertently leaked an early preview of its Jarvis AI on the Chrome Web Store, offering a glimpse into an AI capable of executing tasks autonomously within a web browser. Described as a "helpful companion," Jarvis can reportedly handle routine actions such as grocery shopping, booking flights, and conducting web searches—all without human intervention.

However, the prototype lacked full functionality; a reporter’s attempt to operate it was restricted by access permissions. The Chrome Web Store page for Jarvis was swiftly removed, but its planned December release suggests Google intends for Jarvis to compete with advanced AI assistants like Anthropic's Claude, which can type text, click buttons, and control the cursor on a computer. Read more.

KLING AI video generator allows users to train their own AI video characters

KLING AI now includes a “Custom Models” feature, enabling users to create and train video characters for consistent use across diverse scenes. Testers report that users can upload 10 to 30 video clips, each at least 10 seconds, for training, utilizing video footage rather than static images.

Once trained, the character model can be tagged in prompts to ensure uniformity in videos, including 360-degree views, time-shifted facial angles, and varied camera perspectives. KLING version 1.5 also supports 1080p HD videos in multiple formats, such as landscape, portrait, and square. Additionally, the “Motion Brush” tool lets users specify precise movement for individual elements within scenes. Read more.

AI Development Platforms

• Nvidia announces new robot AI learning and simulation tools

• Flux 1.1 Pro AI image model adds "amateur" RAW photo mode and 4K image generation

• SambaNova and Hugging Face make AI chatbot deployment easier with one-click integration

• Puppygraph speeds up LLMs' access to graph data insights

• Mistral AI takes on OpenAI with new moderation API, tackling harmful content in 11 languages

AI Consumer Products

• Apple iOS 18.2 public beta arrives with new AI features, but some remain waitlisted

• Baidu Readies AI Smart Glasses to Rival Meta’s Ray-Bans 

• Google's Vids AI video maker is rolling out to most Workspace tiers

• Duolingo's new eyerolling emo chatbot Lily briefly replaces CEO on investor call to showcase its AI technology

• ByteDance’s AI can make your photos act out movie scenes — but is it too real?

• New AI experiences for Paint and Notepad begin rolling out to Windows Insiders

AI Robotics and Hardware

• Seattle startup shrinks computer vision AI to fit in your pocket

• Meet ‘Iron’: Made-in-China humanoid robot with 200 degrees of freedom

• AI-driven mobile robots team up to tackle chemical synthesis

AI Policy, Regs, & Legal Developments

• Trump plans to dismantle Biden AI safeguards after victory

• OpenAI’s data scraping wins big as Raw Story’s copyright lawsuit dismissed by NY court

• ‘Unrestricted’ AI group Nous Research launches first chatbot

AI Startups, Acquisitions, and Investments

• The other election night winner: Perplexity

• OpenAI acquires Chat.com

• Amazon mulls new multi-billion dollar investment in Anthropic, the Information reports 

• AI-powered parenting is here and a16z is ready to back it 

• Anthropic teams up with Palantir and AWS to sell AI to defense customers

📄 How Far is Video Generation from World Model: A Physical Law Perspective

📄 LLaMo: Large Language Model-based Molecular Graph Assistant ​​

📄 Controlling Language and Diffusion Models by Transporting Activations  

📄 OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models

📄 ReCapture: Generative Video Camera Controls for User-Provided Videos using Masked Video Fine-Tuning

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!