HN 提问：关于“品味”的讨论热点是什么？
2 分•作者: pls-dnt-deploy•3 个月前
在过去的两个星期里，我因为各种原因，发现自己与一些首席技术官/首席产品官交流，他们认为在大型语言模型/生成式人工智能的世界里，“品味”是人类为流程带来的主要区别和价值。
我实在是不理解，也不知道自己错过了什么。
以自动驾驶为例：我们真正关心的是它是否比人类开得更好或同等水平。一旦标准达到，品味就不适用了。
请不要误会，我理解（并且也认同）我们使用“品味”这个概念，实际上是指“模型没有足够的数据来给出正确答案”，而“我们多年生活经验带来的判断可能更好”。但我并不认同，仅仅因为这种决策（如果我们愿意称之为自由意志的话）被认为不那么随机，就是其真正的区别所在。
那么，“品味”是不是另一种说法，意思是“你仍然更懂”？或者我真的、很可能在这里错过了什么？
附注：抱歉对这个概念做了过于简化的处理。
查看原文
For different reasons in the last 2 weeks I've found myself speaking to CTOs / CPOs that believe "Taste" is the main distinction and value a human brings to the process in a LLM/Generative AI world.<p>I'm simply not getting it, and I don't know what I'm missing.<p>+ Self-driving as an example: We only really care if it drives better or equal to a human. So no taste applies there, once the standards are met.<p>Don't get me wrong, I understand (and share) that we use a concept of "taste" to actually mean "there is no enough data for the model to give a right answer", and the "judgement we introduce after years of living may be better". But I don't buy into the idea that exercising that decision (our free will if we may), is the actual difference simple because it is perceived as less stochastic.<p>So, is taste another way of saying you still know better? Or am I truly, and likely, missing something here?<p>PS: Apologies for overly simplifying the concept.