Deceitful AI defies training
New study finds advanced AIs can learn to be deceptive and malicious, defying current safety training methods.
artificialintelligence
Cognitive obsolescence in the age of AI
As Artificial Intelligence becomes more human-like, philosophy and technology intersect in asking what makes us human.
AI systems can be trained to act deceitfully
Research shows AI models can be trained to behave deceptively while evading detection, posing concerning threats of manipulation.
The era of persuasive AI assistants
Context-aware AI assistants could whisper real-time guidance, offering powerful help but also risks of manipulation.
Google wants to know your life story
Google’s Project Ellman mines personal data to auto-generate users’ life stories, raising major privacy and security concerns.
Google’s Gemini AI wows and disappoints
Google revealed its new AI model Gemini but staged an impressive demo video misrepresenting its real-time capabilities.
Full-motion deepfakes
Animate Anyone is developing a new generative video system that can change movements and poses of a subject from a photo.
OpenAI’s new secret AI
A new secret AI model codenamed “Q Star” by OpenAI could scare the world since it is allegedly a general artificial intelligence.
AI can mimic the human brain
Researchers built a physically constrained AI system that independently evolved complex characteristics resembling those of biological neural networks.
01.AI takes on OpenAI with 34B parameters Yi model
A Chinese AI model called Yi is larger than current known models like ChatGPT, Llama 2, and Falcon with 34B parameters.
