Your Digital Twin, Proxy

• Your personal digital clone for low value tasks

• Gets smarter as you give it commands to learn

• The first truly general AI Agent

Meet your Proxy here

Today:

• OpenAI Launches ChatGPT Desktop Apps

• Google Launches Gemini iPhone App

• Snowflake Unveils New Cortex AI

• Shengshu Launches Image-to-Video Vidu

• François Chollet Exits Google

OpenAI Launches ChatGPT Desktop Apps

OpenAI has released desktop versions of ChatGPT for Mac and Windows to make it easier for people to use the AI in their daily work. The Mac app now lets users open other programs like coding tools directly from ChatGPT. This helps developers who often copy and paste code between apps. 

OpenAI plans to add more programs and bring these features to Windows soon. The Windows app also includes new voice features and the ability to take screenshots, making interactions richer and simpler. These updates aim to keep users within the ChatGPT environment, similar to how GitHub Copilot works with coding tools.

Google Launches Gemini iPhone App

Google has released the Gemini app for iPhone, a chatbot that offers a simple and straightforward interface for users to interact with. The app includes a new feature called Gemini Live, providing a more interactive and conversational chat mode similar to ChatGPT's voice mode.

Gemini app aims to increase user engagement by making it easy to access the chatbot with just one tap, further solidifying Google's presence in the AI chatbot world. While Gemini has limitations on changing settings or accessing other apps on the phone, it can interact with various Google apps, enhancing user experience and accessibility.

Snowflake Unveils New Cortex AI

Snowflake unveiled four major updates to enhance data use for businesses. They introduced Cortex AI tools, which help create chat-based AI applications using company data. Snowflake Intelligence allows users to build "data agents" that answer questions from both organized and messy data and perform tasks in other apps like Salesforce. 

The company also launched Open Catalog for better data organization and Document AI to extract information from documents. To strengthen security, Snowflake added features to protect user accounts and monitor threats. These advancements aim to make data management and AI integration simpler and more effective for enterprises.

Shengshu Launches Image-to-Video Vidu

Chinese startup Shengshu Technology has launched Vidu, an AI tool that turns multiple images into videos. For example, Vidu can combine a shirt, a person, and a moped into a video showing someone wearing the shirt and riding the moped. This feature sets Vidu apart by ensuring the video looks consistent and realistic. 

Vidu became popular on TikTok for creating lifelike videos from just two photos. The tool is already earning money from advertisers and businesses, with customers paying between $13,871 and $138,711 monthly. Shengshu ensures user safety by banning inappropriate content and protecting personal photos according to data laws.

François Chollet Exits Google

François Chollet, a leading expert in artificial intelligence, is leaving Google after almost ten years. He announced on the social media platform X that he will start a new company with a friend but did not share many details. Chollet is best known for creating Keras, a simple tool that helps developers build AI models. Keras is used by over two million people and powers technologies like self-driving cars and recommendation systems on YouTube and Netflix. 

Chollet believes that to make AI as smart as humans, it needs to think more like people do. He will continue to support Keras from outside Google.

🧠RESEARCH

Large Language Models Can Self-Improve in Long-context Reasoning

Researchers developed a method for large language models to improve their reasoning over long texts without needing human help. They generate multiple answers, evaluate them, and use the best ones to train the model further. Tests showed this approach works better than previous methods, enhancing the models' performance.

Motion Control for Enhanced Complex Action Video Generation

Researchers introduced MVideo, a new system that creates longer videos with detailed and smooth actions. Instead of just using text descriptions, MVideo adds motion outlines to guide video creation. This makes the actions clearer and more accurate. Tests show MVideo outperforms previous methods, enabling more complex and dynamic videos.

EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Video Generation

Researchers developed EgoVid-5M, a large collection of five million first-person video clips with detailed action descriptions. This dataset helps create realistic videos from a person’s view, improving virtual reality and gaming. They also built EgoDreamer, a tool that uses these videos to generate new, dynamic first-person videos, enhancing video creation capabilities.

EgoVid-5M: A Large-Scale Video-Action Dataset for Egocentric Video Generation

Researchers developed EgoVid-5M, a large collection of five million first-person video clips with detailed action descriptions. This dataset helps create realistic videos from a person’s view, improving virtual reality and gaming. They also built EgoDreamer, a tool that uses these videos to generate new, dynamic first-person videos, enhancing video creation capabilities.

KMM: Key Frame Mask Mamba for Extended Motion Generation

Researchers developed KMM, a system that enhances video generation of human movements. It improves the existing Mamba model by focusing on important actions and better matching text instructions. KMM reduces errors and uses fewer resources, outperforming earlier methods. This advancement benefits areas like gaming, video creation, and robotics.

🛠️TOP TOOLS

Pear AI - The Open Source AI Code Editor

Kick - Accounting software that does the work for you

Munch - AI video repurposing platform

Learn About Google - Grasp new topics and deepen your understanding with a conversational learning companion that adapts to your unique curiosity and learning goals.

Segwise - AI monitoring of campaign data to help you catch issues before they drain your ad budget.

📲SOCIAL MEDIA

🗞️MORE NEWS

• OpenAI’s ChatGPT for Mac now reads code from apps like VS Code and Xcode, eliminating the need to copy-paste. Although it can’t write code directly, this update enhances developers’ workflows and advances integrated AI systems.

• Google is introducing AI-based security for Pixel phones to fight $1 trillion yearly fraud. Features like Scam Detection warn users about fake calls instantly while keeping their data private. This enhances protection against sophisticated cyber threats.

• Dan Hendrycks, a top expert in making AI safe and advisor to Elon Musk’s AI company xAI, has taken a new job at Scale AI, a major $14 billion business. This shows increased attention to safety in AI development.

• Anthropic partners with the Department of Energy to test its AI, Claude, ensuring it can’t be used to create weapons. Experts try to breach the system, enhancing AI safety for national security as government collaborations grow.

• California’s Diablo Canyon nuclear plant will use AI software to organize millions of old documents and help plan maintenance. This technology aids in meeting new safety regulations, making managing the aging facility more efficient and effective.