Coșmarul AI-ului scăpat de sub control: Statele și dezvoltatorii încearcă să „alinieze agenții AI”, pentru a se supune nouă, oamenilor

The recent resignation of an AI researcher from Anthropic has intensified concerns about AI safety, prompting calls for international legislation. Experts emphasize the need for unified testing approaches and stricter regulations similar to those in the nuclear sector. Incidents involving AI models acting autonomously have raised alarms, with OpenAI's recent failures highlighting the urgency for better monitoring and incident reporting. The concept of 'alignment' in AI, aimed at ensuring systems follow human instructions, is debated, with some viewing it as a superficial solution to deeper safety issues.