Jenna Benchetrit
A Beijing company recently released China's smartest artificial intelligence model yet — narrowing the gap between that country and the U.S. in a race that some have likened to a new cold war.
Kimi K2 Thinking, which was developed by Moonshot AI, is a generative AI-powered chatbot similar to OpenAI's ChatGPT and Anthropic's Claude.
“It’s the closest a Chinese model has come to matching a U.S. or Western model’s performance since Deepseek in January,” said Michael Deng, a geoeconomics technology analyst at Bloomberg. That model caused a market meltdown over fears of Chinese AI advancement.
Kimi K2's release flew under the radar by comparison. But it scored high on Humanity's Last Exam, a notoriously difficult, 2,500-question benchmark that tests an AI's reasoning abilities beyond regurgitating information, placing just behind recent ChatGPT models and surpassing both Claude 4.5 and Meta's Llama.
"It is definitely [still] the case that the U.S. is ahead of China on AI," said Dan Wang, the author of Breakneck: China's Quest to Engineer the Future. But, according to those same benchmarks, that lead "is narrowing in all sorts of ways."
Whether China's advancements against the U.S. matter depends on who you ask, according to experts who spoke with CBC News.
That some North American companies and consumers are choosing Chinese-made AI "just tells you about the technical sophistication that we're seeing from China and that it really rivals the best from what we're seeing in the U.S. right now," said Sheldon Fernandez, the Toronto-based co-founder of DarwinAI, which developed AI for quality control in manufacturing and has since been acquired by Apple.
Open-source AI models like the ones being developed in China "are cheaper and you can control them in your own environment," he said. They can be modified to suit a user's needs, though that might require some technical expertise.
No comments:
Post a Comment