The ChatGPT Observer Edition 29

If you're short on time, here's your one-minute summary of this week's key AI news and advancements:

?? Top Story: Microsoft unveils Copilot Wave 2, introducing Copilot Pages for collaborative work, enhancements across Microsoft 365 apps, and new Copilot Agents with Agent Builder for task automation.

?? ChatGPT & OpenAI News: OpenAI increases rate limits for o1 models; releases safety and security update; Sam Altman steps down from internal Safety Committee; company hires Leah Belsky to lead education expansion.

?? Corporate AI Developments: GitHub integrates OpenAI o1 into Copilot; YouTube introduces AI tools for creators; Alibaba launches Qwen 2.5 models; Runway partners with Lionsgate; Amazon launches AI video generator and Project Amelia.

?? AI Innovations: World Labs secures $230M funding for Large World Models; Qwen 2.5 released as major update to open-source language models with specialised versions for coding and mathematics.

?? Applications of AI: Italy pilots AI-assisted teaching in schools; AI coding assistants impact junior developers' roles; Amazon uses generative AI for same-day deliveries; AI assistant Joannabot reviews iPhone 16.

? Responsible AI: Google to flag AI-generated images in search; California passes 8 new AI laws; LinkedIn opts users into AI training; Guardrails AI framework introduced for secure LLM deployments.

?? AI Learning Hub: New article on Amazon's 750,000+ robot fleet; prompt engineering faces changes with OpenAI's latest o1 model.

?? For more details, read the full newsletter!

?? Top Story: Microsoft Copilot Wave 2

Copilot is the new UI for AI

Last week Microsoft has unveiled a significant update to its Microsoft 365 Copilot, known as "Wave 2," which introduces a range of new features designed to enhance productivity and collaboration through AI. This update includes the launch of Copilot Pages, improvements across various Microsoft 365 applications, and the introduction of Copilot Agents and Agent Builder.

Copilot Pages: A New Era of Collaborative Work

One of the standout features of Wave 2 is Copilot Pages, a dynamic, persistent canvas that facilitates multiplayer AI collaboration. This feature allows teams to work together in real-time, transforming AI-generated content into durable, editable, and shareable work artifacts. Users can collaborate with Copilot as if it were a team member, pulling in information from organisational data, files, and the web to enhance collaborative efforts.

Enhancements Across Microsoft 365 Applications

• Excel: Copilot in Excel is now generally available, introducing support for Python to enable advanced data analysis without coding. It also supports more formulas like XLOOKUP and SUMIF, improving users' ability to handle both numerical and text data.

• PowerPoint: The new Narrative Builder helps users create first drafts of presentations quickly while maintaining creative control. The upcoming Brand Manager feature will allow users to leverage company-branded templates.

• Outlook: The "Prioritize my Inbox" feature analyses inbox content based on user roles, automatically prioritising critical emails and generating concise summaries.
• Teams: Copilot now reviews both meeting transcripts and chat conversations to ensure no questions or actions are missed during meetings.
• OneDrive: Users can now compare up to five documents at once and receive AI-generated insights.
• Word: The interface for creating content has been improved, allowing users to reference emails and meeting notes when drafting documents.

Introducing Copilot Agents and Agent Builder

A major highlight of the Wave 2 release is the introduction of Copilot Agents, which are AI assistants designed to automate and execute business processes. These agents can handle tasks ranging from simple prompt-and-response interactions to more advanced workflows. They operate in the background, managed by Copilot, ensuring security and compliance with Microsoft's Responsible AI principles. To make it easier for users to create these agents, Microsoft has launched the Agent Builder, powered by Copilot Studio. This tool allows users of any skill level to build custom agents directly within BizChat or SharePoint. Users can create agents by simply describing tasks they want automated, enabling rapid deployment without needing coding skills. These agents can be integrated into Teams or Outlook, where they function like a knowledgeable teammate, providing real-time responses and insights

The Future of Work with Microsoft 365 Copilot

The Wave 2 release represents a significant step forward in Microsoft's vision for AI-assisted work. By enhancing existing applications and introducing powerful new tools like Copilot Agents and Agent Builder, Microsoft aims to streamline business processes, reduce costs, and boost team efficiency. As these features roll out over the coming weeks, organisations will have new opportunities to leverage AI for improved productivity and collaboration in the digital workspace.

Watch the announcement:

Read more: Copilot Wave 2 | ZDNET

?? ChatGPT & OpenAI News

• Increased Rate Limits for OpenAI o1 Users: OpenAI is raising the rate limits for Plus and Team users of its o1 models! The o1-mini rate limit is now 50 messages per day, up from 50 per week – a 7x increase! For o1-preview, limits have gone up from 30 to 50 messages per week, helping users explore more without breaking the bank. Read more: OpenAI on X
• OpenAI's Latest Safety and Security Update: OpenAI has released an update highlighting its safety and security practices. The company emphasizes a commitment to user privacy, responsible AI development, and enhanced safety measures for its models. Key areas include data protection, rigorous testing for harmful behaviour, and collaboration with external experts to ensure responsible AI deployment. Read more: OpenAI
• Sam Altman Steps Down from OpenAI’s Safety Committee Amid Governance Shift: OpenAI CEO Sam Altman is stepping down from the company's internal Safety and Security Committee, following recommendations to establish independent oversight for safety measures in AI development. The committee, formed to evaluate OpenAI’s safeguards, emphasized the need for external governance. It will now be chaired by Zico Kolter of Carnegie Mellon University, with other members including Quora’s CEO Adam D'Angelo and retired U.S. Army General Paul Nakasone. This shift highlights OpenAI’s response to balancing innovation with ethical AI governance amid growing pressures from employees and former board members calling for stronger external regulation. Read more: TIME
• OpenAI Hires Former Coursera CRO Leah Belsky to Lead Education Expansion: OpenAI has brought on Leah Belsky, former chief revenue officer at Coursera, as the company's first general manager of education. Belsky will spearhead efforts to bring OpenAI's AI tools, including ChatGPT Edu, to more schools and classrooms. She will focus on building partnerships with educational institutions and ensuring AI is used responsibly in teaching. The initiative aims to expand AI's role in K-12, higher education, and beyond, with universities like Oxford and Columbia already using OpenAI's educational tools. Read more: EduTech Talks

?? Corporate AI Developments

• GitHub Copilot Introduces OpenAI o1 Integration: GitHub has rolled out OpenAI o1 integration into Copilot, allowing developers to leverage the advanced AI capabilities of o1 models directly within their coding environment. This new integration enhances the power of Copilot, providing smarter code suggestions, better context understanding, and increased efficiency for developers. Read more: GitHub
• YouTube Creators Get AI Boost with DeepMind's Generative AI: DeepMind is introducing generative AI tools to empower YouTube creators. These AI-powered features will help creators generate new content ideas, enhance video production, and improve audience engagement by offering creative support throughout the content creation process. Read more: DeepMind
• YouTube Unveils New AI Tools for Creators in 2024: YouTube has announced its "Made on YouTube 2024" initiative, launching new AI-driven tools to enhance the content creation process. These features include AI-assisted editing, content generation, and enhanced music production, all designed to streamline workflows and help creators focus on storytelling. Read more: Google | YouTube on X
• Alibaba Unveils Qwen 2.5 and Expands AI Push with 100 Open-Source Models: Alibaba is accelerating its AI efforts with the launch of Qwen 2.5, the latest in its family of open-source language models. This release includes Text-to-Video, a model that generates videos from text, Coder, designed to assist developers with coding tasks, and Maths, which solves complex mathematical problems. These models are part of Alibaba’s 100 open-source AI offerings, advancing AI innovation across multiple sectors. Read more: Reuters | Qwen
• Runway Partners with Lionsgate and Launches API for AI-Powered Creativity: Runway has secured a major deal with Lionsgate, marking the first collaboration between an AI provider and a major movie studio. This partnership will bring AI-driven tools into film production, enhancing visual effects and creative workflows. Simultaneously, Runway has launched its API to offer developers easy access to its powerful generative AI models, enabling innovative applications across various industries. Read more: VentureBeat | Runway
• Amazon Launches AI Video Generator for Ads and Project Amelia for Sellers: Amazon has introduced two new AI-driven tools. The video generator, currently exclusive for ad content, helps businesses create high-quality video ads quickly and efficiently. Additionally, Amazon has debuted Project Amelia, an AI assistant for sellers, aimed at optimising store management and providing personalised recommendations to improve sales performance. These launches showcase Amazon’s focus on AI-powered solutions for commercial and seller efficiency. Read more: TechCrunch - Video Generator | TechCrunch - Project Amelia
• Apple Intelligence Now in Public Beta: Here's What It Offers: Apple has launched Apple Intelligence in public beta, offering users a suite of AI-powered tools. These features include enhanced personalisation, smart recommendations, and advanced automation across Apple devices. Users can now test these capabilities and enable them through their device settings, marking Apple's latest push into AI-driven user experiences. Read more: TechCrunch
• Aramco Digital and Groq to Build World's Largest AI Inferencing Data Center in Saudi Arabia: Aramco Digital and Groq have announced significant progress on their plan to construct the world’s largest AI inferencing data center in Saudi Arabia. Following an MoU signed at LEAP, this project aims to boost AI capabilities with massive computing power for advanced inferencing tasks. The center is set to play a key role in advancing AI innovation in the region.Read more: Aramco Digital
• Microsoft Unveils GRIN MoE AI Model for Coding and Math: Microsoft has introduced GRIN-MoE (Gradient-Informed Mixture-of-Experts), a new AI model designed for coding and math tasks. This model uses a Mixture-of-Experts architecture, activating only 6.6 billion of its 16x3.8 billion parameters at a time, offering efficient and high-performance computation. GRIN-MoE scored 79.4 on the MMLUbenchmark and 74.4 on HumanEval, outperforming competitors like Mixtral and LLaMA 3. While excelling in reasoning tasks, it faces challenges in conversational and multilingual AI. Read more: VentureBeat
• Adobe Firefly Unveils Text-to-Video Generative AI Features: Adobe has announced new text-to-video generative AI features in its Firefly platform. These tools allow users to create videos from simple text prompts, enhancing creative workflows for video production. Currently in preview, Firefly aims to make video creation more accessible and streamlined, allowing creators to generate professional-level content with ease. Read more: The Verge
• Salesforce Unveils AgentForce for Building Autonomous AI Bots: Salesforce has introduced AgentForce, a new tool designed to help businesses create autonomous AI bots. These bots can perform a variety of tasks, from customer service to internal operations, streamlining workflows and enhancing efficiency. AgentForce allows companies to deploy AI-driven agents that can operate independently, offering a powerful solution for automation and improving customer experiences. Read more: CIO
• Roblox Launches Generative AI for Instant 3D World Creation: Roblox has unveiled a new generative AI tool that enables users to instantly create 3D environments. This technology allows creators to generate complex virtual spaces with minimal effort, transforming the game-building process by making it faster and more accessible. The AI tool is designed to simplify the creation of immersive worlds, further expanding Roblox's appeal to developers and players alike. Read more: MIT Technology Review
• Workday Acquires AI-Powered Document Platform Evisort: Workday has acquired Evisort, an AI-powered document management platform. Evisort specialises in automating document processing and contract management, using AI to streamline workflows and improve operational efficiency. This acquisition strengthens Workday’s capabilities in AI-driven solutions, particularly in managing legal and business documents, and enhancing its enterprise resource offerings. Read more: TechCrunch
• GitHub Expands Copilot Ecosystem with New Extensions in Public Beta: GitHub has launched Copilot Extensions in public beta, further enhancing its Copilot ecosystem. These extensions allow developers to customise and extend Copilot’s functionality to suit specific workflows and coding needs. With this update, GitHub aims to give developers more control and flexibility over how Copilot assists in code generation and development tasks. Read more: GitHub
• Together.AI and Meta Launch LLaMACoder for AI-Powered Coding: Together.AI and Meta have introduced LLaMACoder, an AI model designed to assist developers with code generation, debugging, and coding suggestions across various languages. A standout feature of LLaMACoder is its ability to generate an entire app from a prompt, streamlining the development process and making it easier for developers to build complex applications with minimal input. Read more: Meta

?? AI Innovations

• World Labs: Advancing AI with Spatial Intelligence: World Labs is a cutting-edge AI company focused on building Large World Models (LWMs) to give AI the ability to perceive, generate, and interact in 3D worlds. Founded by AI pioneers such as Fei-Fei Li, the company aims to go beyond text and 2D image generation, pushing the boundaries of spatial intelligence. By creating AI that can understand 3D space, World Labs hopes to revolutionise fields like design, engineering, and virtual world creation. With $230M in funding from top investors, including Andreessen Horowitz, Adobe Ventures, and NVIDIA, World Labs is developing tools for creative professionals and the broader public to build and interact with unlimited 3D worlds. Read more: World Labs AI
• Qwen 2.5: A Major Leap in Open-Source AI Models: The Qwen team has announced the release of Qwen 2.5, a major update to its family of open-source language models. This release includes the core Qwen2.5 model, along with specialised models for coding (Qwen2.5-Coder) and mathematics (Qwen2.5-Math), available in sizes ranging from 0.5B to 72B parameters. These models have significantly improved knowledge, with superior performance in coding and math tasks, achieving benchmarks such as MMLU: 85+and HumanEval: 85+. Key enhancements include the ability to handle long text generation (up to 8K tokens), better structured data handling, and improved instruction following. The models support over 29 languages, making them versatile across different use cases. Qwen2.5 models are now available via Hugging Face and licensed under Apache 2.0 for most variants, marking one of the largest open-source AI releases in history. Read more: Qwen

?? Applications of AI

• Italy Pilots AI-Assisted Teaching to Boost IT Skills in Schools: Italy is testing AI-assisted teaching tools in schools as part of a national initiative to improve students' IT skills. The project aims to integrate AI into classrooms to enhance learning, with a focus on equipping students with critical tech skills for the future. This pilot could pave the way for broader adoption of AI in education across the country. Read more: Reuters
• AI Coding Assistants: Impact on Junior Developers' Roles: The rise of AI coding assistants is transforming the software development landscape, raising concerns about the future of junior developers. These AI tools are automating many entry-level coding tasks, such as debugging and code generation, potentially reducing the demand for junior developers. As AI continues to evolve, the role of human coders may shift towards more complex problem-solving and higher-level responsibilities. Read more: CIO
• Amazon Leverages Generative AI to Boost Same-Day Deliveries: Amazon is using generative AI to enhance its logistics operations and increase the number of same-day deliveries. By optimizing route planning and inventory management, the AI system improves delivery speed and efficiency. This technology allows Amazon to meet growing customer demand for faster deliveries, especially in urban areas. Read more: CNBC
• iPhone 16 Review by Joannabot: An AI Assistant's Take on Apple's Latest: In an experimental twist, Joanna Stern from The Wall Street Journal handed over part of her iPhone 16 review duties to Joannabot, a generative AI assistant trained on over a decade of her iPhone reviews. Powered by Google’s Gemini Flash AI model, Joannabot answers user questions about the iPhone 16 lineup, offering insights on upgrades, features, and performance. Read more: The Wall Street Journal

? Responsible AI

• Google to Flag AI-Generated Images in Search: Starting later this year, Google will begin flagging AI-generated images in its search results. This move is part of Google's effort to increase transparency and help users distinguish between AI-generated content and real-world images, ensuring more informed browsing and media consumption. Read more: TechCrunch
• California Passes 8 New AI Laws: What's Now Illegal: California has introduced eight new laws aimed at regulating the use of AI. These laws focus on issues such as data privacy, transparency, and preventing AI from causing harm. The regulations prohibit the use of AI in manipulating elections, deepfake creation without consent, and other potentially harmful applications. This marks a significant step in addressing the ethical and legal challenges posed by AI technology. Read more: TechCrunch
• LinkedIn Quietly Opts Users Into AI Training Using Their Data: LinkedIn has automatically opted users into a system that uses their data to train generative AI models, as revealed by 404Media. This update, introduced alongside a revised privacy policy, allows LinkedIn to utilise user data for developing AI tools, such as writing assistants. While users can opt-out by adjusting their privacy settings, data already used for AI training cannot be undone. Importantly, LinkedIn states that personal data from users in the EU, EEA, or Switzerland is not used for training. Read more: The Verge
• Guardrails AI: A Framework for Secure, Reliable LLM Deployments: Guardrails AI is an open-source framework designed to add safety and reliability to AI applications, particularly those using large language models (LLMs). It offers key functionalities like input validation—ensuring user inputs meet specific criteria—and output validation, which checks that AI-generated responses follow defined rules and constraints. Guardrails AI is model-agnostic, allowing flexibility across different LLM providers, and offers low-latency performance for fast, real-time validations. Developers can integrate it via a Python library, API server, or enterprise solution for centralised control. Use cases include protecting user privacy, maintaining compliance, ensuring brand consistency, and fact-checking LLM outputs. The framework is actively supported on GitHub, with community contributions enhancing its validators and components. Read more: Guardrails AI
• Hong Kong Considers AI Regulations for Finance Sector: Hong Kong is preparing to issue its first policy statement on the use of AI in finance, aiming to guide ethical AI applications in trading, investment banking, and cryptocurrencies. The government plans to unveil the framework during Fintech Week in late October, signalling support for AI as local regulators tackle confusion over AI services amidst US-China tensions. The anticipated guidelines will promote responsible AI use, with more specific rules expected in the future.Read more: Yahoo Finance
• Gov. Newsom Signs AI Bills to Regulate Actor Replicas and Deepfakes in Hollywood: Governor Gavin Newsom has signed several AI-related bills aimed at protecting actors' digital likenesses and curbing the spread of deepfakes, particularly in political ads. Key measures include Assembly Bill 2839, which allows candidates to seek legal action against deepfakes that harm their reputation, and AB 2655, which requires tech platforms to label and remove fake content. Additionally, new laws grant actors more control over their digital replicas, particularly after death or if AI is used without proper consent. These regulations respond to concerns raised during Hollywood's recent strikes over AI advancements in entertainment. Read more: LA Times
• Google to Roll Out AI-Generated Image Detection Across Multiple Services: Google is integrating C2PA (Coalition for Content Provenance and Authenticity) technology into its search engine, advertising systems, and potentially YouTube to help detect AI-generated images. C2PA works by attaching a cryptographically protected metadata file, or manifest, to images, showing if they were created or modified by AI. This will be included in Google's "About this image" feature and aid in enforcing ad policies. Google aims to improve transparency around AI-generated content, further building on its previous efforts like SynthID, which watermarks AI-created images. Read more: SiliconANGLE
• TrueMedia: AI Tool for Detecting Political Deepfakes in Social Media: TrueMedia.org offers a free tool that enables users to detect deepfakes in social media posts with industry-leading accuracy. Built for critical election audiences worldwide, the platform aggregates analysis from multiple AI detectors to identify manipulated videos, images, and audio files with approximately 90% accuracy. This solution addresses the growing problem of deepfakes, with a 1740% surge in North America alone last year. Users can upload files or social media links for fast detection results. Read more: TrueMedia

?? AI Learning Hub

• Learn about Amazon's 750,000+ robot fleet and the future of human-robot collaboration.

• Prompt Engineering Faces Changes with OpenAI’s Latest o1 Model: OpenAI's new o1 generative AI model introduces key shifts in prompting and prompt engineering, requiring users to adapt their strategies. Unlike previous models, o1 automatically integrates chain-of-thought reasoning, eliminating the need for explicit step-by-step prompts. Key prompting tips for o1 include focusing on simplicity, using delimiters to separate elements, and streamlining retrieval-augmented generation (RAG) to optimize performance. While o1 excels in narrow domains like coding and mathematics, it’s less effective in broader applications, making it crucial for users to tailor their prompting techniques. Read more: Forbes

??? We'd Love to Hear from You!

Did you find this edition insightful? Like, share your thoughts, and ask us anything.

Stay curious, and until next time, enjoy exploring the vast possibilities AI offers.

???? Subscribe on LinkedIn The ChatGPT Observer Newsletter

As we explore the captivating world of AI technologies, it's important to note that the content of this article is a product of human effort, enhanced by AI tools and additional resources. We have made sure to properly attribute and link all referenced articles to their original sources. For the visuals:

• Images sourced from external sites are credited within the article.
• Other images were generated using DALL-E, MidJourney and IdeoGram with enhancements made in Canva. These AI-generated and enhanced images are designed to visually represent each section's key themes.