Hold onto your lab coats, tech fans! OpenAI just dropped its new o1-Preview AI models – think of them as ChatGPT’s brainy cousins who pause to ‘think’ before answering. Designed to tackle math, coding, and science challenges, these models aim to fix AI’s infamous habit of making up facts (aka ‘hallucinations’).
CEO Sam Altman called it a ‘new paradigm’ for AI reasoning, comparing its problem-solving skills to PhD students in fields like physics and biology. In tests, the models scored an 83% success rate on tough math Olympiad questions – way above GPT-4o’s 13%!
But here’s the catch: Altman admits the tech ‘is still flawed’ and might lose its ‘wow factor’ after extended use. Still, researchers say it hallucinates less – though they’re not claiming victory over fake facts yet.
From helping doctors analyze cells to letting coders build multi-step apps, OpenAI’s brainy bots could soon power next-gen tools. Just don’t ask them to write your thesis… for now.
Reference(s):
cgtn.com