✅ ChatGPT finally does tasks

Hey folks,

A relatively unknown AI lab from China called DeepSeek, has caused a stir in Silicon Valley by releasing AI models that outperform the best from U.S tech giants, despite being developed at a lower cost and using less-powerful hardware. In December, DeepSeek launched a free, open-source LLM that reportedly took just two months and under $6 million to build, utilizing Nvidia's lower-capability H800 chips. The battle for AI dominance between U.S v China is heating up.. let’s dive in.

_{Read Time: 6 minutes} 

✅ OpenAI launches ‘Operator’ to take over your tasks

🐝 The Buzz: OpenAI just dropped a new AI agent named Operator, which can handle tasks like booking travel and making restaurant reservations without needing much input from users. Currently, it’s exclusive to those with the ChatGPT $200 Pro subscription, but OpenAI plans broader access soon.

• The CUA model allows Operator to navigate websites like a human, albeit with built-in limits and user supervision for sensitive tasks. 

• OpenAI is testing Operator primarily on U.S. markets first, while residents of other countries wait their turn.

• OpenAI is in cahoots with companies like DoorDash and Uber to ensure compliance with their services. For security reasons, certain intricate tasks and sensitive actions require user intervention. 

• The Operator isn't perfect and can stumble on complex interfaces, but it’s a promising step towards making automation more human-like.

💡 Takeaway: OpenAI’s launch of Operator is a pivotal moment as it aims to simplify internet tasks for users. By tackling automation with a user-conscious approach, OpenAI is not just pushing tech boundaries, but also easing the digital load on everyday users. Making our lives easier in a secure way? Now, that’s what we like to see!

Pro tip: The key to getting the most out of Operator is being very specific with your prompts. Instead of “find me a restaurant,” try “find me a cheap Italian pet-friendly restaurant in Manhattan with free parking on Friday night”

📝 Anthropic stops hallucinations with AI backed sources

🐝 The Buzz: Anthropic just dropped a new feature called "Citations" for its Claude AI family, designed to help developers ground AI responses in real documents, like emails. Citations allows models to reference specific passages from source documents and is available on Anthropic's API and Google’s Vertex AI platform. It's tailored for tasks like document summarization and customer support.  

• Citations is available only for Claude 3.5 Sonnet and Claude 3.5 Haiku models, not every Claude on the block.

• Developers can use Citations to make their AI more reliable, particularly for applications like Q&A where sourcing information can make all the difference.  

• Anthropic aims to reduce AI-induced hallucinations and errors with this feature, providing users with more accurate information.

💡 Takeaway: The introduction of Citations is a step towards making AI answers more trustworthy, which is something that anyone who’s faced the infamous "hallucinating AI" situation can appreciate. Having grounded responses may not only save time and energy, but it also cultivates a more reliable environment for using AI in everyday tasks, making it more user-friendly and error-resistant.

📰 Like newsletters? Here are some others our readers also enjoy.

💻️ How to create a Wiki-like AI report on any topic

• Go to the STORM Research Platform and create an account.

• Enter your topic in the search bar.
  eg. “AI Agents ability to run tasks”

• Enter the writing purpose 
  eg “Educational document to help understand Agents”

• Your Wiki-like document is created with sources for reference.

• Check the sources and information for accuracy.

🐝 AI buzz bits

👩‍💻 Google is rolling out a groundbreaking feature for Chromebooks that allows users to control their devices using facial expressions, aimed primarily at improving accessibility for those with motor impairments, while also launching new classroom tools.

🤖 Salesforce CEO Marc Benioff warns that today's executives will be the last to manage entirely human workforces, as AI integration transforms job roles, with many companies planning to downsize as they automate tasks by 2030.

🚀 The NVIDIA H200 GPU significantly elevates performance for large language models like DeepSeek V3, boasting enhanced memory capacity and bandwidth, paving the way for advanced inference and training capabilities essential for handling complex, large-scale AI tasks.

🌟 Hugging Face has unveiled two new compact multimodal models, SmolVLM-256M and SmolVLM-500M, optimizing performance while being the smallest Vision Language Models ever, ensuring efficient processing on constrained devices.

🖼️ Generative Art

Industrial design blueprint of an industrial robot arm, detailed drawing with notes and annotations, white background, schematic, top view, front perspective, mechanical engineering concept art.

AI productivity
🔥 Tools to take your workflow to the next level.

1. Reachy - Automation tool for LinkedIn outreach that enhances networking and lead generation while preserving user privacy.

2. Magic Patterns - AI-driven design tool that quickly transforms ideas into UI components for faster product iteration.

3. Hedra - AI-driven platform that generates and styles characters and videos quickly from text, image, or voice inputs.

4. AI Undetect - A powerful tool that rewrites AI-generated text into undetectable human-like content, supporting over 20 languages.