China’s leading artificial intelligence (AI) startups are looking for new opportunities in their field, after preview of advanced new series large language models (LLM’s) – the technology supporting generative AI services – by ChatGPT Creator OpenAIaccording to executives of Alibaba Cloud‘S Apsara Conference in Hangzhou.
OpenAI o1already announced earlier this month by the Microsoft-backed company, represents a so-called generative, pre-trained transformer model, which is designed “to understand complex tasks and solve harder problems than previous models in science, coding, and mathematics,” according to the San Francisco-based company.
“(OpenAI o1) is indeed of great importance,” Yang Zhilinfounder of Moonshot AIsaid Thursday at the main forum of the three-day conference, which runs until Saturday. Alibaba Cloud is the digital technology backbone of Alibaba Group Holdingowner of the South China Morning Post.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP knowledgeour new platform featuring carefully curated content with explanations, FAQs, analyses and infographics, brought to you by our award-winning team.
“The key question here is whether you can scale even further via reinforcement learning, which greatly increases the upper bound of AI,” Yang said, referring to the so-called scaling law that shows how the performance of an LLM improves as the size and training data increase.
Moonshot AI Founder and CEO Yang Zhilin. Photo: Weibo alt=Moonshot AI Founder and CEO Yang Zhilin. Photo: Weibo>
The 31-year-old founder of Moonshot AI expects OpenAI o1 to “bring about some changes in the structure of many industries and create new opportunities for startups.”
According to an update from OpenAI this week, the new models have been trained “to spend more time thinking through problems before responding to them, just as a human would.”
Through training, they can “learn to refine their thinking processes, try different strategies and recognize their mistakes,” the US company said.
“Once (a company) reaches a certain threshold of computing power, it can make fundamental innovations in algorithms and even achieve breakthroughs in fundamental models,” Yang said.
This development, Yang said earlier this weekrepresents a paradigm shift because LLM developers, no longer with organic data to train their models, are now turning to a technique called reinforcement learning to recreate thought processes and thus generate more data.
Following that vision, the AI start-up Step fun‘s founder and CEO, Jiang Daxinsaid that reinforcement learning “has been generalized to a higher level by OpenAI o1”.
Jiang said OpenAI o1 is expected to increase opportunities for greater innovation in fundamental model development and AI applications.
An industry executive addresses the audience during the main forum of Alibaba Cloud’s three-day Apsara conference in Hangzhou, capital of eastern Zhejiang province, on September 19, 2024. Photo: Xinhua alt=An industry executive addresses the audience during the main forum of Alibaba Cloud’s three-day Apsara conference in Hangzhou, capital of eastern Zhejiang province, on September 19, 2024. Photo: Xinhua>
However, he points out that computing power is still a problem for startups. US trade sanctions the availability of advanced restrict semiconductorssuch as that of Nvidiafor AI development projects of Chinese companies.
“The computational power required is still not small, especially if we pursue a generalized reasoning model,” Jiang said. “But if the goal we pursue is artificial general intelligence (AGI), we will stick to it no matter how much it costs.”
Existing AI technologies all operate within a set of predetermined parameters, but AGI involves the development of systems with autonomous self-control, some degree of self-awareness, and the ability to learn new skills.
No more than 10 AI startups – including the four so-called “AI tigers” – In China, investment in reinforcement learning could double, according to a Baichuan AI a conference staff member who wished to remain anonymous.
These AI tigers – Moonshot AI, Baichuan AI, Zhipu AI And MiniMax – are all aimed at LLM development.
This article originally appeared in the South China Morning Post (SCMP)the most authoritative voice covering China and Asia for over a century. For more SCMP stories, explore the SCMP app or visit the SCMP’s Facebook And Twitter pages. Copyright © 2024 South China Morning Post Publishers Ltd. All rights reserved.
Copyright (c) 2024. South China Morning Post Publishers Ltd. All rights reserved.