Good morning. It’s Friday, October 4th.

Did you know: On this day in 1957, the Soviet Union launched Sputnik 1 into orbit.

In today’s email:

• New “Canvas’ ChatGPT Tool

• OpenAI Funding Round

• Meta Glasses Dox in Real Time

• 5 New AI Tools

• Latest AI Research Papers

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

OpenAI launches new 'Canvas' ChatGPT interface tailored to writing and coding projects

OpenAI's latest "canvas" interface adds a new dimension to ChatGPT, designed for hands-on writing and coding tasks. This feature pops open a dedicated workspace next to the usual chat window, letting users generate and tweak content without the hassle of re-prompting.

Rolling out first to Plus and Teams users, it’s a clear response to rivals like Anthropic’s Artifacts, which have embraced editable AI workspaces. In "canvas," users can refine emails, adjust text length, or switch languages with simple commands. For coders, it’s even more seamless—inline comments, explanations, and code reviews make debugging less of a chore.

OpenAI is positioning "canvas" as a practical tool for tackling larger projects in pieces, rather than forcing everything through one prompt. After beta, free users will also get a taste, underscoring OpenAI’s aim to keep pace with competitors while enhancing the creative workflow. Read more.

OpenAI closes funding at $157 billion valuation, as Microsoft, Nvidia, SoftBank join round

OpenAI has raised $6.6 billion in new funding, boosting its valuation to $157 billion. Thrive Capital led the round, with participation from Microsoft, Nvidia, and SoftBank, indicating strong confidence in the AI leader. This funding, which follows a previous valuation of $80 billion, will support frontier AI research and enhance computational capabilities.

A significant aspect of this funding is a $4 billion revolving credit line, providing OpenAI with financial flexibility as it pursues growth. Despite reporting losses exceeding $5 billion this year, projections suggest revenues could rise to $11.6 billion by 2025, reflecting strong investor confidence. Read more.

Related Story: OpenAI asks investors to avoid five AI startups including Ilya Sutskever's SSI, sources say

College students used Meta’s smart glasses to dox people in real time

A demonstration by two Harvard students reveals the risks associated with Ray-Ban Meta smart glasses, showcasing their potential for real-time doxxing. The project, named I-XRAY, utilizes the glasses' livestreaming feature on Instagram to capture faces, which are then matched with public databases to obtain personal information such as names, addresses, and phone numbers.

The demo illustrated how the students identified classmates and engaged strangers based on this data. While the creators stress that I-XRAY is not meant for misuse and will not be released, it highlights significant privacy concerns surrounding facial recognition technology in consumer devices. The discreet design of these glasses complicates public awareness of filming, raising echoes of the backlash that Google Glass faced. Read more.

Quick Hits

Google brings ads to its AI Overviews, Lens search tools: For mobile users in the U.S., relevant ads now appear beneath the Google AI-generated content. Additionally, Google’s image recognition tool, Lens, has been upgraded to enhance shopping-related queries. Users can now access product reviews, pricing, and store information alongside ads. Voice commands also allow users to refine their searches, while Lens can now explain objects in videos. Read more.

A co-lead on Sora, OpenAI's video generator, has left for Google: Tim Brooks, a co-lead on OpenAI’s video generator Sora, has joined Google DeepMind to work on video generation technologies and “world simulators.” In a post on 𝕏, he expressed gratitude for his tenure at OpenAI, where he was instrumental in launching Sora in January 2023, overseeing its research direction and large model training.

Google's lightweight Gemini 1.5 Flash-8B hits general availability: The model features a substantial 1 million-token context window and operates at 40% faster speeds than OpenAI's GPT-3.5 Turbo, while providing a token input price that appeals to developers. Following initial access for select users and a free trial via the Gemini mobile app, the model is now accessible through the Gemini API and Google AI Studio. Enhancements were made based on developer feedback, with the 8B version nearly matching the original in performance across critical benchmarks.

• Cohere just made it way easier for companies to create their own AI language models

• Google developing new AI model to rival OpenAI o1 in reasoning and complex problem solving

• AI researchers discover "Law of the Weakest Link" in language models

• GPT-o1-mini helps mathematicians with complex proofs, but it's complicated

• ‘In awe’: scientists impressed by latest ChatGPT model o1

• Accenture forms Nvidia-focused AI unit with 30,000 employees

• Oura Ring 4 debuts slimmer design, additional sizes, and improved sensing starting at $349

• Nvidia newest open source model benchmarks

• OpenAI CEO Sam Altman outlines vision for next-generation AI hardware and interfaces 

• Black Forest Labs, the startup behind Grok's image generator, releases an API

• TikTok’s parent launched a web scraper that's gobbling up the world’s online data 25-times faster than OpenAI

• DreamWorld playtest of AI text-to-3D-asset generation coming to Steam

• Inside Elon Musk’s AI party at OpenAI’s old headquarters

📄 Atlas-Chat: Adapting Large Language Models for Low-Resource Moroccan Arabic Dialect

📄 From Code to Correctness: Closing the Last Mile of Code Generation with Hierarchical Debugging ​​

📄 Video Instruction Tuning With Synthetic Data

📄 RATIONALYST: Pre-training Process-Supervision for Improving Reasoning  

📄 LEOPARD : A Vision Language Model For Text-Rich Multi-Image Tasks

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!