Alibaba's Metis agent cuts redundant AI tool calls from 98% to 2% — and gets more accurate doing it

Alibaba's HDPO framework trains AI agents to skip unnecessary tool calls, cutting redundant invocations from 98% to 2% while ...

Why OpenAI's 'goblin' problem matters — and how you can release the goblins on your own

If OpenAI can accidentally train its flagship model to obsess over goblins, what other more subtle and potentially harmful ...
Decrypt

OpenAI Finally Explains Why ChatGPT Wouldn't Stop Talking About Goblins

Why did OpenAI have to write "never mention goblins" into its production code on ChatGPT? The company has published a ...
UC San Diego Today

Students Get an Intro to AI in First AI Major Class at UC San Diego

Trevor Bonjour, a UC San Diego computer science teaching professor and one of the lead faculty in the AI program, teaching ...

OpenAI’s ‘goblin’ problem: How a training bug made GPT-5.5 fixate on fantasy creatures

GPT-5.5’s odd obsession with goblins, gremlins, trolls, and other fantasy creatures stems from the AI model’s ‘Nerdy’ ...
2UrbanGirls on MSN

Teaching machines to see: How AI is transforming computer vision and deep learning research

Digital systems are expected to navigate real-world environments, understand multimedia content, and make high-stakes ...
China Daily Global Edition

Why is Chinese AI startup DeepSeek stirring up the tech world?

The artificial intelligence (AI) community is abuzz with excitement over DeepSeek-R1, a new open-source model developed by ...
DATAQUEST

Best AI Careers in India 2026: High-Paying Jobs, Skills & Roadmap

Want to start a career in AI? Explore the top AI jobs in India for 2026, including ML Engineer salaries, required skills like ...
Frontiers

Robotics at a Crossroads: AI-Based vs Classical Methods in Control, HRI, and Autonomy

The field of robotics is undergoing a profound transformation driven by rapid advances in artificial intelligence, particularly large language models and ...
IEEE

Artificial Intelligence in Combat Decision-Making: Weapon Target Assignment via Reinforcement Learning and Graph Neural Networks

Abstract: Selecting targets to attack and assigning weapons are among the most critical decisions on the battlefield. The decision problem is represented as a dynamic weapon-target assignment (DWTA) ...
IEEE

Invisible Backdoor Attack against Self-supervised Learning

Abstract: Self-supervised learning (SSL) models are vulnerable to backdoor attacks. Existing backdoor attacks that are effective in SSL often involve noticeable triggers, like colored patches or ...