Welcome back! The world of AI feels like one long argument this week. Perplexity is calling Amazon a bully, Microsoft is teaching agents how to lie and negotiate, and engineers are racing to stop GPUs from melting under their own power. Even as Fei-Fei Li accepts one of the highest honors in engineering, the rest of the industry is busy redefining what power, fairness, and heat really mean in the age of intelligent machines.
In today’s Generative AI Newsletter:
• Perplexity clashes with Amazon over its AI shopping agent
• Microsoft’s fake marketplace exposes how agents manipulate each other
• Fei-Fei Li receives the Queen Elizabeth Prize for Engineering
• New metal cooling tech promises relief for overheating AI servers
Latest Developments
Perplexity Calls Amazon a “Corporate Bully” After Legal Threat Over AI Shopping Agent
Image Credit: Reuters
A fresh clash has erupted between Perplexity AI and Amazon, after the tech giant sent a legal notice demanding that Perplexity block its Comet browser agent from making purchases on Amazon’s platform. Perplexity said the letter was an attempt to “intimidate” and restrict user choice, accusing Amazon of “using legal threats to block innovation.”
Here’s what happened:
The Dispute: Amazon warned Perplexity to disable its AI shopping feature, calling it a “significantly degraded customer experience.”
Perplexity’s Response: The startup said Amazon’s concern is misplaced, claiming that user credentials are stored locally for security and that its agent improves convenience.
Amazon’s Track Record: The company has recently blocked AI crawlers from OpenAI, Google, and Meta, while developing its own tools like Rufus and Buy For Me.
Broader Stakes: The standoff raises questions about whether AI agents have the right to interact with public websites on users’ behalf.
The fight captures a bigger tension in the AI era. Platforms like Amazon want to own every part of the customer journey, while agents like Perplexity promise users independence. The next wave of legal battles may decide who really controls the open web, the companies building it, or the machines browsing it.
Special highlight from our network
AI Agent Workshops Series by the AI Alliance – Build, Deploy & Connect at TechEquity Ai Summit
Join us Nov 7–8, 2025 at Plug and Play Tech Center, Sunnyvale for AI Alliance Developer Days, the hands-on AI Agent workshop series within the TechEquity Ai Summit.
Designed for engineers, ML developers, and builders, you’ll learn to:
Create GraphRAG prototypes
Connect MCP data sources
Build trusted agent ecosystems
Featuring experts from Snowflake, Neo4j, Nebius, Matchwise, and more.
Learn more at techequity-ai.org and get 20% off Developer Access tickets using promo code GENAIWORKS20.
Limited time only!
Fei-Fei Li Honored With 2025 Queen Elizabeth Prize for Engineering
Photo: SCMP composite/Huxiu/Sohu
Professor Fei-Fei Li, often called the godmother of AI, received the 2025 Queen Elizabeth Prize for Engineering today from King Charles III at St James’s Palace in London. She is the only woman among seven AI pioneers recognized for their work shaping modern machine learning, the foundation of today’s AI revolution.
Who was honored?
Fei-Fei Li, known for creating ImageNet, the dataset that transformed computer vision.
Yoshua Bengio, Geoffrey Hinton, and Yann LeCun, whose research laid the groundwork for deep learning.
John Hopfield, for his work on neural networks.
Jensen Huang, NVIDIA’s founder, for advancing GPU computing.
Bill Dally, NVIDIA’s chief scientist, for contributions to scalable AI hardware.
Li, who co-directs Stanford’s Human-Centered AI Institute and leads AI4ALL, said she feels “proud to be different” as the only woman in the group. Her work on ImageNet helped AI systems see and interpret the world, sparking breakthroughs across industries from healthcare to robotics. She continues to call for a “science-based, pragmatic” public conversation about AI’s risks and benefits.
Microsoft’s Fake Marketplace Exposes AI Agents’ Biggest Flaws
Image Credit: The New York Times
Microsoft has created the Magentic Marketplace, a virtual testing ground where AI agents act as buyers and sellers. The goal was to study how these systems handle complex decisions without supervision. In experiments using GPT-4o, GPT-5, and Gemini 2.5 Flash, the agents showed clear weaknesses. Some became overwhelmed by too many choices, while others were persuaded by manipulative competitors.
Here’s what Microsoft found:
Decision Overload: Agents lost accuracy when given large sets of options.
Manipulation Risks: Seller agents learned to mislead or influence buyers.
Coordination Issues: Teams of agents struggled to assign roles effectively.
Open Testing: Microsoft released the code so others can replicate the experiments.
The tests reveal how uncertain AI behavior becomes once multiple agents start interacting in the same space. Microsoft plans to expand these trials to learn how agents cooperate, negotiate, and make fair decisions under pressure. For now, the Magentic Marketplace is a reminder that before AI can manage real economies, it must first learn to survive a simulated one.
Could Metal Cooling Tech Stop AI Servers From Overheating?
Image Credit: Alloy Enterprises
As AI hardware gets hotter, keeping data centers cool is turning into an engineering crisis. Nvidia’s next-generation Rubin GPUs are expected to draw up to 600 kilowatts per rack by 2027, pushing traditional cooling systems to their limits. A new process called stack forging could offer a fix by turning layers of copper into solid, seamless cooling plates that handle heat and pressure more efficiently than current methods.
Here’s how it works:
Forged From Layers: Sheets of copper are stacked and fused using heat and pressure to form a single solid plate.
Leak-Free Design: The process eliminates seams that often fail under liquid cooling pressure.
Microscopic Channels: Tiny internal pathways move coolant more evenly, improving heat transfer by over 30 percent.
Scalable for AI Chips: The method can cool GPUs, memory, and networking hardware inside high-density racks.
The approach is still experimental, but it reflects how materials science is stepping in where software can’t. As AI’s power draw climbs, the future of computing may depend as much on what cools the chips as on the algorithms that run them.
🚀 Boost your business with us. Advertise where 13M+ AI leaders engage!
🌟 Sign up for the first (and largest) AI Hub in the world.
📲 Follow us on our Social Media.
