Key insights

• 🚀 🚀 Significant AI updates this week include a knowledge cut-off for ChatGPT extended to June 2024, enhancing its ability to provide current information.
• 📸 📸 ChatGPT gains improved image understanding, making it more adept at handling visual content and enhancing STEM subject performance.
• 🌍 🌍 New regional features are being rolled out for users in Switzerland, Iceland, and Liechtenstein, offering tailored experiences in these areas.
• 🛠️ 🛠️ The Canvas feature updates allow users to build and test HTML code directly, improving accessibility for non-coders and streamlining web project development.
• 🤖 🤖 Two new open-source AI models from China, focusing on large context windows and advanced video recognition, have emerged, showcasing local capabilities.
• ⚠️ ⚠️ Medi's new memories feature automatically collects user data, highlighting ongoing privacy concerns and implications for user consent.
• 📜 📜 The U.S. copyright office clarifies that purely AI-generated content is not copyrightable, impacting how digital content is created and protected.
• ✨ ✨ Despite advancements, existing models like ChatGPT continue to outperform new contenders, particularly in user experience and task handling.

Q&A

• What did the U.S. copyright office say about AI-generated content? 🇺🇸

  The U.S. copyright office has clarified that purely AI-generated content cannot be copyrighted. It noted that while unique prompting styles do not render AI outputs as unique art, editing or combining AI outputs with other works may still be subject to traditional copyright laws. New AI models like Europe’s Mistral 24B and Tulu Free 405B are emerging, but they may not yet offer significant benefits to consumers.

• What privacy concerns have arisen from Medi's new feature? 🤔

  Medi's new memories feature automatically collects user data without providing users an opt-out option, raising significant privacy concerns. Users are required to share personal data for the assistant to customize recommendations, incorporating information from sources like Instagram and Facebook profiles, which heightens the risk of data breaches and misuse.

• What are the latest Chinese AI models and how do they compare to ChatGPT? 🧠

  Two new open-source AI models from China have emerged; one features a compact model suitable for local use with a 1 million context window, while the other is a vision model excelling in video recognition. Despite their advancements, existing models like ChatGPT still outperform them, particularly in terms of overall capabilities, although the new models offer promising benchmarks and efficient context retention.

• What new applications for ChatGPT have been introduced? 💻

  Recent updates have introduced new features for writing and displaying HTML and React components through ChatGPT. A government-specific version has also been released to address privacy concerns, and there's been an integration of Deep Seek into the AI search capabilities of Perplexity, which enhances functionality for Pro users. Additionally, new open-source models from China aim to provide expanded context windows and vision capabilities.

• How has the Canvas feature in ChatGPT improved coding capabilities? 🚀

  The latest upgrades to ChatGPT's Canvas feature allow users to view, build, and test HTML code for web projects directly within the interface, making it much user-friendlier, especially for non-coders. The OAI model 01 can now be enabled for larger projects, enabling quick iteration and previews, which streamlines the coding experience further.

• What are the key updates for ChatGPT this week? 🧠

  This week, ChatGPT has seen major updates including a knowledge update extending to June 2024, enhanced image understanding capabilities, improved performance in STEM subjects, and the roll-out of new regional features for users in Switzerland, Iceland, and Liechtenstein. Additionally, a new 'think button' has been introduced for better user experience, along with the anticipated release of O Free Mini, which promises to be faster and more efficient.

• 00:00 This week, there are significant AI updates, with a focus on Chinese releases and enhancements in ChatGPT, including a knowledge update to June 2024, improved image understanding, and new regional features. 🧠
• 02:37 The latest upgrades to the Canvas feature in ChatGPT enhance coding capabilities, allowing users to view, build, and test HTML code for web projects directly within the interface, making it more user-friendly for non-coders. 🚀
• 05:27 Recent updates highlight ChatGPT's enhanced coding capabilities in HTML and React, the introduction of a government-specific version for privacy-focused applications, and advancements in AI search engines with perplexity's integration of deep seek, along with new open-source models from China. 💻
• 08:03 Two new open-source AI models from China emerge, featuring a compact model with a large context window suitable for local use and a vision model excelling in video recognition, but existing models like ChatGPT remain superior. 🧠
• 10:43 Medi has launched a new memories feature that automatically collects user data without opt-out options, raising privacy concerns, while Deep Seek AI released an open-source image generation model that currently produces unsatisfactory results. 🤔
• 13:22 The U.S. copyright office clarified that purely AI-generated content is not copyrightable, noting that editing or combining AI outputs with other works may fall under traditional copyright. Additionally, new AI models are emerging, although they may not significantly benefit consumers yet. 🇺🇸