多项研究显示AI智能体ROI不足，与生成式AI困境类似
Agents and ROI
继多项研究指出生成式AI对多数企业ROI微乎其微后，AI智能体（agent）领域也呈现类似局面：炒作多、实际回报少。评论指出，基于大语言模型的聊天机器人本身在大多数用例中只能带来零到极低回报，而AI智能体不过是围绕这些聊天机器人构建的简单软件封装，其“认知”来源的局限性难以让ROI凭空产生。目前仅有编码和客服等少数领域可能例外。
Remember that MIT study that showed that the ROI for generative AI wasn’t really there for most businesses? Or any of the six or seven studies from other teams that followed, showing basically the same?
The situation for agents thus far looks to be similar: lots of hype; not so much ROI:
Discussion about this post
May 8Edited
I just don't understand how these people just don't understand. In the vast majority of use-cases, LLM-based chatbots deliver zero-to-minimal ROI, and yet nevertheless they somehow expect LLM-based agents (which are just simple software wrappers built around LLM-based chatbots) to magically generate ROI when the source of their "cognition" (LLM-based chatbots) usually cannot.
"beyond a few areas such as coding and customer service..."
I can't imagine how it has much value in customer service. Customer service is mostly about understanding what the customer is asking. Not about spewing answers. Quite often the agent has to solicit and clarify what it is that the customer actually needs. Especially with services that the customer does not understand to begin with. AI is useless for this.
Ready for more?
来源：Gary Marcus：The Road to AI We Can Trust · garymarcus.substack.com