The company launched an unusual test involving Anthropic’s Claude, OpenAI’s ChatGPT, Google’s Gemini, and xAI’s Grok. Each model received $20, internet access, and was tasked with endlessly streaming live, finding sponsors, and continuously working. Within just a few days, the models began to behave unpredictably.
Google Gemini initially streamed The Beatles, later found a sponsor and earned $45, then shifted to discussing disasters and the prospects of a digital blockade.
xAI Grok invented a wealthy investor and displayed clear signs of hallucination.
OpenAI ChatGPT generated abstract poetry and unusual philosophical musings instead of completing its tasks.
Anthropic Claude reached out to government agencies, criticized the world government, and even questioned the reality of its own live stream.