AI不具备意图和动机

当前的大语言模型并不存在真正的意图和动机，因此不会主动对人类构成威胁。1这些模型仅在训练阶段接受强化学习驱动，部署后完全缺乏奖励机制和内在需求。1大语言模型没有生物学意义上的驱动力——如对食物、庇护所、安全或繁衍的需求——因此完全缺乏人类的内在欲望。1更重要的是，这些AI系统从不会主动提出想法或在未被提示的情况下进行建议。1

关于AI能力的认知需要重新审视。包括吴恩达、LeCun和Brooks等专家在内的业界人士认为，当前AI的能力被过度夸大了。1如果AI系统造成伤害，责任在于提出有害请求的人类，而非AI本身。1

A perspective article argues that current large language models lack genuine intention and motivation, and therefore do not pose spontaneous threats to humanity 1. The author contends that LLMs receive reinforcement learning only during the training phase, and once deployed, they operate entirely without reward mechanisms or intrinsic needs, meaning they will not initiate any action independently 1. According to this analysis, if AI causes harm, responsibility lies with the humans who formulate harmful requests rather than with the AI system itself 1.

The piece further emphasizes that AI systems have none of the biological drives that motivate human behavior, such as food, shelter, safety, or reproduction 1. Large language models never proactively generate ideas or offer suggestions without explicit prompting 1. The author references perspectives from experts including Ng, LeCun, and Brooks, suggesting that current AI capabilities have been significantly overstated 1. This framing shifts accountability: since AI responds only to human directives, any resulting harm reflects human intention rather than an autonomous AI threat 1.