OpenAI’s latest AI models are getting better and better, but they hallucinate significantly more often than their predecessors. In certain tests, the GPT o3 and o4-mini “reasoning” systems achieved error rates of up to 79% – a trend that leaves even developers puzzled.
OpenAI’s latest AI models are getting better and better, but they hallucinate significantly more often than their predecessors. In certain tests, the GPT o3 and o4-mini “reasoning” systems achieved error rates of up to 79% – a trend that leaves even developers puzzled.
Shuhei Yoshida, former head of Sony’s gaming division, comments on rising game prices. He describes the price development as inevitable in view of growing development costs and inflation. At the same time, he advocates AA games with smaller teams.
