Did DeepSeek Steal from OpenAI?

OpenAI has accused Chinese AI start-up DeepSeek of using its proprietary models to train a competing open-source model, potentially violating intellectual property rights. OpenAI suspects DeepSeek employed "distillation," a technique that transfers knowledge from larger models to smaller ones, which, if done externally, breaches OpenAI’s terms of service.

DeepSeek’s recent AI model release, built at a fraction of OpenAI’s cost, has surprised the industry. Some experts suspect it was trained using OpenAI’s GPT-4 outputs. OpenAI and Microsoft previously blocked suspected DeepSeek accounts for violating API terms.

The controversy underscores ongoing challenges in protecting AI intellectual property, with AI companies frequently using commercial model outputs to train their own systems. OpenAI continues to work on countermeasures and has called for US government cooperation to safeguard its technology. Meanwhile, OpenAI faces its own copyright lawsuits from publishers and authors accusing it of unauthorized use of their content.

Link to full article

Another Chinese Model Qwen claims to be better than OpenAI and DeepSeek

Alibaba has released a new version of its AI model, Qwen 2.5-Max, claiming it outperforms OpenAI’s GPT-4o, DeepSeek’s DeepSeek-V3, and Meta’s Llama-3.1-405B. The launch comes amid rising competition from DeepSeek, a Chinese AI startup that has disrupted the market with low-cost, high-performance models, causing a shake-up in Silicon Valley and driving down tech stocks.

DeepSeek's rise has also triggered a wave of updates from Chinese competitors, including ByteDance and Baidu. DeepSeek’s previous model, V2, sparked a price war, forcing Alibaba and others to slash AI pricing. The startup’s founder, Liang Wenfeng, downplays price wars, emphasizing AGI (Artificial General Intelligence) as the company’s primary goal. Unlike major Chinese tech firms, DeepSeek operates with a lean, research-focused team.

The timing of Alibaba’s release coinciding with the Lunar New Year suggests urgency to respond to DeepSeek’s rapid ascent, which has put pressure on both international and domestic AI players.

Link to full article

DeepSeek New Image Model Janus Pro vs OpenAI Dalle, Midjourney and Stable Diffusion

DeepSeek, a Chinese AI company, has released Janus-Pro-7B, a new image generator that it claims outperforms OpenAI’s DALL-E 3 and Stability AI’s Stable Diffusion in benchmark tests. However, independent verification of these claims is still pending.

The release of Janus-Pro-7B coincided with a major stock market reaction, as Nvidia’s stock lost $589 billion in value in a single day. The Janus Pro models, available on GitHub and Hugging Face, come in 1B and 7B parameter versions. If benchmarking results hold up, DeepSeek’s approach could pose a challenge to leading AI image generators.

Currently, Janus-Pro-7B has limitations, such as image uploads restricted to 384 x 384 pixels, though it can generate 768 x 768 images in demos. DeepSeek’s chatbot has also faced performance issues due to cyberattacks, leading to temporary restrictions on new user registrations. Despite these challenges, DeepSeek’s advancements are being closely watched in the AI industry.

Link to full article

ChatGPT Gov for US Government

OpenAI believes AI adoption by the U.S. government can enhance efficiency, productivity, and global leadership. To support this, they have launched ChatGPT Gov, a version of ChatGPT tailored for government agencies, deployable via Microsoft Azure to ensure security, privacy, and compliance with stringent cybersecurity standards. ChatGPT Gov offers features similar to ChatGPT Enterprise, including GPT-4o, shared workspaces, custom GPTs, and an administrative console.

Since 2024, over 90,000 users across 3,500 U.S. agencies have exchanged 18 million+ messages using ChatGPT for tasks like research, translation, and administrative support. Early adopters include the Air Force Research Laboratory, Los Alamos National Laboratory, and the State of Minnesota, among others. OpenAI aims to expand ChatGPT Gov and strengthen AI integration in government operations while maintaining public trust.

Link to full article

How to build a Web AI Agent

Excited to share my latest project: Web AI Agent—an AI-powered assistant that autonomously browses the internet, retrieves information, and summarizes results for you. Imagine having your own digital research assistant that works while you sip your coffee! ☕🤖

🔹 Key Features:
✅ Opens a web browser & searches the internet
✅ Extracts relevant information
✅ Summarizes results concisely

This project is part of Day 6 of my "How to Build an AI Agent" series, where we turn theoretical concepts into real AI applications. No complex infrastructure—just lean, efficient code!

💡 Tech Stack:
🔹 LLM: OpenAI GPT-4o (switching to an open-source LLM soon!)
🔹 Agent Framework: Microsoft AutoGen

Want to try it out? The complete code is on GitHub! 🚀🔗 GitHub Repository
Link to details on how to build it