Let's talk about ChatGPT, it's really hard!

Preface

Since I used ChatGPT at work, my work efficiency has skyrocketed. Some people may say that Google is enough, but Google mainly searches for existing keywords on the Internet based on the keywords you enter. The results are matched and sorted. There are three problems.

If your keywords are not so accurate, then Google will not be able to do anything.

Even if you hit the keyword, Google’s search The result may not be what you want, because it is not what you want in terms of overall semantics. It is like if you type "I love China", you may use Google, and it will find that the key to "China" is hit. Words, the result shows you a bunch of information about what China is like, but if you use ChatGPT, it can guess your semantics, and then reply to you with words such as "It seems that you are a good young man and will be a pillar in the future." Such a comparison clearly shows the difference, precisely because Google is based on keywords, and ChatGPT is based on semantics, so as long as your sentence is longer, Google will be more inaccurate, while ChatGPT will be able to guess your semantics more accurately (provided that you The long sentence description given is clear enough)

Even if the results given by Google hit your semantics, the search results may also be wrong, because Google ranks based on the weight of the web page. If this web page If it has a high weight, its ranking will be high, but the content of the web page itself may be wrong. Just because many people visit it, its weight ranking will be high, but the answer itself may be wrong. . If you use ChatGPT, such errors will most likely be avoided. We can think that it will be a robot with certain thinking ability and will avoid similar problems. I am deeply impressed by this. For example, I asked a lot of questions about ChatGPT. The sample codes given are all bug-free, and using Google, sometimes those web pages that rank highly may also have bugs.

It is precisely because ChatGPT can guess your semantics that it can "make something out of nothing", that is, even if your keywords are not hit on the Internet, it can still give answers that are not available on the Internet, like an independent Thinking people, but Google is different. If it is not available on the Internet, it is not available. It can only be searched based on the keywords you gave, but it does not match your semantics.

Let me share two examples used in my work to experience the power of ChatGPT

case 1: Find "js request short link to get the corresponding long link"

Search using Google is as follows

Using ChatGPT the results are as follows:

You can clearly see the difference. Google gives the implementation principle of short links, while ChatGPT not only understands my semantics (pay close attention and you will find that I mistakenly wrote how to if) but also accurately gives the corresponding response. The code implementation completely met my requirements

case 2: "There are two labels lined up in a row. One label has a fixed width, and the other label has an adaptive width. If the text part exceeds, Use an ellipse to represent it, please use css to represent another label"

The answer given by Google is as follows:

One answer was this.

It does solve the "need to use ellipsis after the text exceeds the limit", but it does not meet my other requirement: adaptive width, We can see that this answer limits the width to 100px. This is what we mentioned at the beginning that Google is not good at processing long sentences, and it cannot understand your semantics

Next, let’s take a look at the performance of ChatGPT.

It can be seen that it not only completely guessed my semantics, but also gave the bugfree code, including the naming of classes in the code, and comments, which are the ceiling of standard code! Very developer friendly!

Comparing the two to judge the difference

The above are just two cases where we use ChatGPT in coding. In fact, our entire team has already used ChatGPT heavily in our work, such as using It analyzes popular video scripts and then writes scripts with similar styles. For another example, we ask it to write articles with similar styles to other popular public accounts. Originally, these tasks required recruiting corresponding operators to do it, but with ChatGPT Finally, all these tasks can be done by it, and as far as I know, other companies on the market are currently developing some intelligent customer service, paper checking, mock interview and other applications based on ChatGPT. The application scenarios are extremely wide, and the subversion it brings to the industry is enough to cause Our emphasis

I wrote a paragraph in the circle of friends before: "Being able to use ChatGPT proficiently for work will be a required item in the JD (job description) of the engineer position soon" because based on my experience of using it. In other words, the productivity brought by using it is really great. If you use it skillfully, it may be worth several ordinary engineers. As shown on the cover, it has passed the test of Google L3 engineers, and the salary of this engineer is equivalent to 18w. This means that it is likely to replace the work of ordinary engineers in the future.

Of course, ChatGPT is just a tool after all, and it cannot completely replace humans in the short term. However, we still have to embrace changes in time, make full use of it to improve our work efficiency, and hand over some repetitive labor to it. What we have to do is to write good code, but also think deeply about the business and become experts in the industry. ChatGPT cannot replace humans in this part.

The above is the detailed content of Let's talk about ChatGPT, it's really hard!. For more information, please follow other related articles on the PHP Chinese website!