告诉HN：人工智能实验室由没有全球制造经验的人领导
我意识到，一些顶尖的美国人工智能实验室由缺乏制造专业知识的人领导，他们对现实世界中的事物是如何制作的并没有深入了解。我并不怀疑他们作为世界级研究者的经验，但研究实验室并不代表真实世界，真实世界中存在复杂的动态。
他们所有的“模型”都是在寻找解决方案的问题，但这并不意味着这些模型对某些任务没有用处。
在发布模型之前，他们从未停下来问自己：“我们在朝什么方向努力，或者我们在解决什么问题？”因此，才会出现那些毫无意义的基准测试。
这让我想起一句话：“如果你不理解一个过程，你就一无所知。”
即使他们现在所称的“人工智能”能够编写软件，这也表明他们认为软件等同于代码行或功能，忘记了软件是一个或多个个体在对世界观进行交流和妥协后产生的复杂或简单过程的结果。
而在人工智能领域，可能只有Yann LeCun对这个世界有真正的把握。
查看原文
I have realized - some of the premier american 'a.i' labs are headed by people without manufacturing expertise - of | or how things are made in the real world. I'm not doubting their experience as world class researchers - but the research lab is not the real world - with complex dynamics.<p>all their 'models' are solutions seeking problems. but it doesn't mean the models are not useful for some tasks.<p>before releasing models - they have never stopped to ask - what are we working towards or what are we solving. hence the nonsense benchmarks.<p>it reminds me of a certain quote "if you don't understand a process, you don't know shit".<p>even what they propose as 'a.i' now writes software - shows they think software == lines of code or features. forgetting software is the result out of complicated/simple process of 1 or more individuals talking/compromising on a world view.<p>& probably the only person in 'a.i' who has grip on the world in Yann LeCun.