How do we solve problems in the age of AI?

Let's take a look at a problem. This is a typical "chickens and rabbits in the same cage" problem, which is a must-know for students. When learning about equations in junior high, this is a typical question for two equations. And when it comes to computation, it's a typical programming problem. Let's see how we can use AI to solve it.

Do you want to see the explanation? Click below, and there it is.

In the above example, we solve problems using everyday natural language, rather than code that requires specialized training.

Now, let's look at a slightly more complex problem: investment. Investment is a relatively complex task, requiring the collection of a large amount of data and consideration of various factors before making a decision. Let's see how AI can assist us in making investment decisions.

This AI's response may not be perfect, but what we can be sure of is that AI is already on its way to more usable and perfect. It has set off; it's just a matter of time.

The recent performance of Claude 3 has been jaw-dropping: it demonstrates an understanding and fluency on complex tasks that are close to human level, leading the forefront of general intelligence, and has even reached a doctoral level in the field of scientific research. Let's think it twice. If a person achieved the levels of the doctoral research level in every field, (nothing is not included), and how powerful that person would be, how powerful this kind of deep understanding and integration (interdisciplinary, cross-domain) would make.

AI models are incredibly powerful, and for most of our applications, we only need to use them through their interface. This interface is very simple, which is just our natural language.

The Best Programming Language is Natural Language

The development of science is a process of liberating human physical and mental constraints. Fifty years ago, the abacus was very popular because, with this tool, people could calculate quickly. Back then, the best accountants were those who could move beads of abacus the fastest. In the 1990s, in China, accounting students had to take accounting grade exams, where a set of arithmetic problems were given, and grades were determined based on the speed of calculation with an abacus. At that time, the abacus was the most effective resource, and abacus skills were even more popular than algorithm programming skills today. Subsequently, with the computerization of accounting, that is, the use of electronic computers to replace manual bookkeeping, calculating, totaling, etc., the idea of computerized accounting was proposed much earlier but due to the scarcity of computers at the time, it was not until the late 1990s that the skills for computerized accounting became widespread. Afterward, the abacus only appear in a computation museum, or from a personal hobby rather than a professional necessity. In that context, accounting software, Excel, and other computer tools became essential skills. This was followed by computer operation certifications. Later on, with the development of computer hardware and the advent of the internet, anyone could learn programming languages to write software on personal computers, making software development an essential skill. People learned various programming languages to complete different types of software programming work. Professional software programming became a useful skill. The best tool for utilizing information resources was various software tools and programming skills that required professional learning.

Returning to our subject, with the current state of technological advancement, what is the best tool for us to utilize information and computing resources? For most people, in most cases, natural language might be the best tool for operating information and computing resources. As seen in the previous examples, the current development of AI can already help us complete most information processing tasks, although it may not yet be perfect. We can imagine that current large models, only 2-3 years old, can already provide us convenient services. It's not hard to imagine that, given time and thousands of times more computing power, we will undoubtedly be able to use natural language descriptions to have powerful artificial intelligence systems, invisible to us, but can do more things for us.

It is foreseeable that, when artificial intelligence development matures, any user of natural language will be able to describe their needs, the logic behind their requirements, and the desired output. The clearer the description, the more likely the results will meet one's needs.

Just as in the automobile era, most time people do not need to learn how to choose horses, how to feed horses, nor how to repair cars; what is needed is learning how to drive.