Chinese AI start-up founder hails ‘important improvement’ in OpenAI’s newest model
LLM developers have run out of organic data to train models, so they are turning to reinforcement learning to generate more data
The development of artificial intelligence (AI) has seen a paradigm shift with the launch of OpenAI’s latest 01 model, according to the founder of one of China’s biggest generative AI start-ups, as Chinese firms race to catch up with their US peers.
An “important improvement” made by OpenAI o1, the US start-up’s newest large language model (LLM) released last week, is that it can accomplish tasks that would typically take humans a long time to think about and to complete, Yang Zhilin, founder and chief executive of Chinese unicorn Moonshot AI, said in a sharing session on Saturday in Tianjin, according to a transcript published by Tencent News.
The development represents a paradigm shift because LLM developers, which have run out of organic data to train their models, are now turning to a technique known as reinforcement learning to recreate thought processes and in turn generate more data, Yang said.
San Francisco-based OpenAI, which launched OpenAI o1 on September 12, said in a blog post that its large-scale reinforcement learning algorithm “teaches the model how to think productively using its chain of thought”, and outperforms its other LLMs in reasoning-heavy tasks in fields that include science, coding and maths.
The OpenAI o1 model “spends more time thinking through problems before they respond, much like a person would”, according to OpenAI.
