纽约大学教授加里·马库斯表示，美中人工智能竞赛将以平局告终，而不会是任何一方的决定性胜利。
The US-China artificial intelligence race will end in a tie rather than a decisive victory for either side, New York University professor Gary Marcus said.
他将这场竞争比作可口可乐和百事可乐，两家汽水制造商各自保留市场份额，却没有将对方挤出市场。他说，美国的人工智能模型总体上更好，不过中国的模型更便宜。
He likened the competition to Coke and Pepsi, where both soda makers each retain their market share without driving the other out of business. US AI models are generally better, he said, though the ones from China are cheaper.
他在新德里举行的彭博创新经济论坛上接受彭博电视台伊冯·曼采访时表示：“我们错误中的悲剧在于，这两个国家都在付出如此巨大的努力，试图赢得一场我认为无法赢得的战争。因此，他们把事情推向了零和博弈。”他重申了此前对人工智能风险的警告，并以随之而来的安全漏洞为例，称他关于人工智能编程代理将成为“安全噩梦”的预测已被证明是正确的。
“The tragedy of our error is that these two countries are putting in so much effort to try to win what I think is an unwinnable war,” he said in a Bloomberg TV interview with Yvonne Man at the Bloomberg New Economy Forum in New Delhi. “And so they’re pushing things into being a zero-sum game.”He reiterated his earlier warnings about the risk of AI, citing the security failures that followed, saying his prediction of AI coding agents becoming “a security nightmare” had proven correct.
在监管方面，马库斯呼吁在推出新模型之前进行外部审查，类似于美国食品药品监督管理局进行的审查，他表示，独立评估人员将在模型发布前审查其效益是否大于风险。他还呼吁美中两国加强合作，而不是禁止对方的人工智能模型。
On regulation, Marcus called for external reviews ahead of new models, similar to those conducted by the US Food and Drug Administration, where independent assessors examine whether a model’s benefits outweigh its risks before release, he said. He also called for more cooperation between the US and China, rather than banning AI models from another country.
他说：“关于安全的规则应该是普遍的，我真的认为美中两国应该就此达成协议。”两国可以在事件和错误方面保持沟通，“在这里单打独干并不是一个好主意。”
“The rules around safety should be universal, and I really think that US and China should make a deal around that,” he said, where they could be in communication about incidents and errors. “Going it alone here is not a great idea.”