AI News

New research from OpenAI and MIT Media Lab shows that intensive ChatGPT use correlates with higher loneliness and lower socialization. Voice mode helps in the short term, but for vulnerable users it may create a vicious cycle of dependency.

This week delivered several defining moments in artificial intelligence. OpenAI launched its GPT-5.5 model, Anthropic is pushing Claude into personal life, and the European Commission is stepping up pressure on Google. Here is your essential roundup of everything that matters.

More than half of South Koreans now use ChatGPT for search, leaving traditional portals scrambling as conversational AI becomes the go-to channel for discovering information. What does this shift mean for the Czech Republic?

Large language models are no longer standalone applications. Google has embedded Gemini directly into Chrome's address bar, Apple has opened Siri to ChatGPT and Gemini, and Korean Kakao has launched ChatGPT inside its largest communication platform. While the world obsesses over the parameters of new models, the real battle is over who gets AI into the hands of the most users. And it seems closed ecosystems are being left behind.

The new AI agent GOLF.AI CONCIERGE automates customer support and bookings at golf clubs. Featuring the voice of legend Sir Nick Faldo, it delivers a personalized experience with no need for human staff.

OpenAI's detective hunt for ChatGPT's unusual behavior uncovered a deeper problem in model training. What looks like an amusing bug reveals weaknesses in the reinforcement learning mechanism and the danger of 'universal transfer' of errors into base models.

A new term that emerged in early 2025 describes a software development approach where programmers "go with the flow" and let AI write code for them. What exactly is vibe coding, who is using it, and what risks does it bring?

WHO warns the world will be short 11 million healthcare workers by 2030. Google DeepMind's answer: AI co-clinician — a multimodal agent that sees, hears, and speaks with patients in real time. How trials at Harvard and Stanford are shaping the future of medicine, and what it means for healthcare in the Czech Republic.

Perplexity AI processes 780 million queries monthly with instant source citations. It works across EU languages, offers a free tier and premium plans up to €189/month. How does it compare to Google and ChatGPT for European users?

Experts say prompt engineering certificates are worthless. The era of agentic systems and "swarms" is here instead. Discover why you must learn to build autonomous systems, not just write instructions.