Daily Productive Sharing 1010 - Building AI Products

One is to treat it as a science problem - this is early, and the models will get better.

Daily Productive Sharing 1010 - Building AI Products
Photo by Jepretualang / Unsplash

One helpful tip per day:)

What is the biggest problem with current AI products? Benedict Evans offers his insights:

  1. General-Purpose Interface: Most current AI products present themselves as general-purpose interfaces. However, the answers they provide are not always 100% accurate, which can cause significant confusion for users. This approach is different from the software design of the past 50 years, where each piece of software was designed as a specialized tool to help users accomplish a specific task.
  2. Specialized Tools Analogy: Think about tools like drills, washing machines, and blenders. Fundamentally, they all use a motor to drive different components, but people rarely buy a standalone motor to solve various problems like drilling holes, washing clothes, or blending food. This is exactly the issue with current AI products.
  3. Improving AI Performance: While AI model performance will continue to improve, this alone will not solve the fundamental problem of how AI products are designed and used.
  4. Focused AI Products: One solution is to create AI products that are designed to solve specific domain problems. By focusing on particular areas, these products can offer more reliable and accurate solutions.
  5. Embedding AI into Other Products: Another approach is to embed AI models into other products. This is similar to what Apple demonstrated at their recent WWDC, where AI capabilities were seamlessly integrated into their existing product ecosystem.

If you enjoy today's sharing, why not subscribe

Need a superb CV, please try our CV Consultation

当下的 AI 产品最大的问题是什么?我们又该如何改进?Benedict Evans 给出了自己的见解:

  1. 当下的 AI 产品大多把自己开放成一个通用的界面,而提供的答案又不是100%准确,所以会给用户造成很大的困扰。这和过去50年的软件设计完全不一样 -- 每一个软件都是包装成一个专用工具,引导用户做好一件事;
  2. 想象一下,电钻,洗衣机,搅拌机的本质都是一个会转动的电机带动不同的组件,但很少有人会去买一个单独的电机然后来解决钻洞,洗衣服或者搅拌的问题。这恰恰就是目前 AI 产品的问题;
  3. AI 模型性能当然会继续提升,但这仍旧无法解决上述问题;
  4. 一种解决思路是把 AI 产品做成只解决特定领域问题的产品;
  5. 另一种就是把这些模型嵌入到其他产品中,就像 Apple 在最近的 WWDC 中展示的那样。

如果你喜欢的话,不妨转发给身边的朋友 ⬇️