Good morning. It’s Monday, November 11th.

Did you know: Meta has a career page specifically for veterans?

In today’s email:

• OpenAI Exploring Alternate Strategies?

• Altman: AGI in 2025

• AI Art Sells for $1M

• Google’s ReCapture

• 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 KAPA AI}

Kapa turns your knowledge base into a reliable and production-ready Large Language Model

Join +100 leading companies like OpenAI, Docker, monday.com, Grafana, CircleCI, Silicon Labs, Tomtom and Mapbox in using kapa to improve developer engagement, finding documentation gaps, and reducing repetitive support inquiries.

Kapa turns your knowledge base into a reliable and production-ready LLM-powered AI assistant that answers technical product questions instantly and helps improve your documentation.

Here is why +100 leading companies trust Kapa:

• Explainable answers: Minimal hallucinations with answers grounded in source materials, rigorously tested across 100+ deployments.
• Comprehensive knowledge integrations: The platform offers 30+ data source integrations (e.g., website crawlers, Slack channel readers, wiki/forum parsers) that automatically refresh, ensuring up-to-date information.
• Off-the-shelf deployments: 1-click deployments including website chatbot widgets, Slack bots, Discord bots, Zendesk assistants, and APIs for easy integration.
• Continuous improvement: Built-in features and analytics help find gaps in your documentation, adjust AI behavior, and enhance answer quality over time.
• Enterprise-grade security: SOC II Type II certified with advanced PII detection/masking technology for data security and privacy.

Add Ask AI to your docs like Docker by booking a demo on kapa.ai!

OpenAI reportedly developing new strategies to deal with AI improvement slowdown

OpenAI’s upcoming model, codenamed "Orion," shows signs of decelerated performance gains compared to past versions, according to a new report in The Information. While Orion outperforms current models overall, it lacks the transformative leap seen from GPT-3 to GPT-4, particularly in specific tasks like coding.

To counter this deceleration, OpenAI has formed a foundations team tasked with enhancing model improvement strategies, especially as the pool of available training data narrows. Key strategies include leveraging synthetic data generated by AI models and refining models post-training. The shift highlights cautious approach in addressing technical challenges and setting new precedents in managing AI development plateaus. Read more.

Sam Altman on AGI's Arrival in 2025 and What’s Next for OpenAI 

In a recent Y Combinator "How to Build the Future" episode, Sam Altman, CEO of OpenAI, expressed his anticipation for the arrival of Artificial General Intelligence (AGI) in 2025. When asked by YC President Garry Tan what excites him most about the coming year, Altman pointed to AGI as a defining milestone.

He also shared reflections on OpenAI's evolution and offered guidance for founders navigating the shifting AI landscape. Alongside this, Altman disclosed that he is expecting a child in 2025.

Robot artist Ai-Da's portrait of Alan Turing shatters auction records selling for over $1 million—a first for AI artwork

In a milestone for AI-generated art, Ai-Da, a humanoid robot artist, made history by selling her portrait of Alan Turing for over $1 million at Sotheby’s Digital Art Sale. Named "A.I. God," the 2.2-meter portrait surpassed pre-sale estimates and sparked discussions around AI's place in art. Ai-Da, designed to resemble a human, can communicate through AI and was conceptualized by art specialist Aidan Meller.

Using advanced AI, Ai-Da creates original artwork, guiding her artistic decisions based on conversations and visual analyses. The Turing portrait, in muted tones with fragmented facial elements, subtly addresses ethical questions about AI's evolution, echoing concerns once voiced by Turing himself.

Meller notes that Ai-Da’s work, described as “ethereal and haunting,” challenges viewers to contemplate the “god-like” power of AI and its implications for the future. Read more.

Google's ReCapture lets users change camera angles in existing videos 

Google's ReCapture introduces a novel AI-driven method enabling users to retroactively adjust camera angles in pre-recorded videos. Employing a two-step process, ReCapture first generates an “anchor video” with the desired camera movements. This anchor, while preliminary, guides the model but may initially display inconsistencies. To refine it, the system uses masked video fine-tuning that incorporates both temporal and spatial Low-Rank Adaptation (LoRA) layers.

Temporal LoRA handles movement continuity, while spatial LoRA preserves detail. Relying on video-to-video translation, ReCapture utilizes diffusion models like CAT3D, allowing smooth panning, tilting, and zooming adjustments without requiring full retraining. Though still experimental, ReCapture demonstrates a forward leap in user-friendly, professional-grade video manipulation. Read more.

• 𝕏 is testing a free version of AI chatbot Grok

• OpenAI loses another lead safety researcher, Lilian Weng

• ChatGPT hits 3.7 billion visits in October as growth accelerates to 115% year-over-year

• ByteDance adds video generator to China’s most popular AI chatbot Doubao

• AI Detected Her Early Breast Cancer After Annual Test Came Back as Normal: ‘I feel so lucky’

• China develops first AI robot lifeguard to stand 24-hour watch over river

• FrontierMath Launches New Benchmark with Unpublished Problems, Top LLM Scores Just 2%

• OpenAI's ChatGPT election query redirects highlight concerns over AI's factual reliability

• a16z VC Martin Casado explains why so many AI regulations are so wrong

• Waymo’s new AI research shows “promising” self-driving gains

Writer RAG tool: build production-ready RAG apps in minutes

RAG in just a few lines of code? We’ve launched a predefined RAG tool on our developer platform, making it easy to bring your data into a Knowledge Graph and interact with it with AI. With a single API call, writer LLMs will intelligently call the RAG tool to chat with your data.

Integrated into Writer’s full-stack platform, it eliminates the need for complex vendor RAG setups, making it quick to build scalable, highly accurate AI workflows just by passing a graph ID of your data as a parameter to your RAG tool.

Learn more about our production ready RAG tooling here.

📄 Artificial Intelligence, Scientific Discovery, and Product Innovation

📄 Mixture-of-Transformers: A Sparse and Scalable Architecture for Multi-Modal Foundation Models ​​

📄 StdGEN: Semantic-Decomposed 3D Character Generation from Single Images

📄 DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusion

📄 TIP-I2V: A Million-Scale Real Text and Image Prompt Dataset for Image-to-Video Generation

Thank you for reading today’s edition.

(and happy Veteran’s Day)

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!