Author: Liudahou

This morning, I was awakened by a string of messages in the group chat, saying that AI is going to replace humans again. The last time I heard this was when AlphaGo defeated humans in Go, but in the past few years, AI seems to have not made much progress.

So I opened it to take a look, and couldn’t help but shudder. Although the interface is even more rudimentary than that of a chatbot a few years ago, it seems to be capable of doing something that is beyond our imagination.

It can directly write several pages of dialogue. Someone asked it to write script lines with actors from Friends, someone asked it to write code, and someone even tried using it to write this post (not me).

It can be said that it is no longer a simple chatbot and can even have some profound exchanges with you. Even its former boss, Elon Musk, was frightened and exclaimed “everyone is deep into ChatGPT”.

So, what is ChatGPT that drives people crazy? Let it answer for itself:

OpenAI officially claims that ChatGPT was created and trained with the help of humans. Human trainers ranked and rated its query answers in its early versions. Then, this information was fed back into the system, and the system would adjust the answers based on trainers’ preferences – this is a standard method of training artificial intelligence known as reinforcement learning.

OpenAI was co-founded in 2015 by Musk, Altman, and others. Later, Musk withdrew and Microsoft invested 1 billion dollars. OpenAI first released a large model called GP-3 with trillions of parameters, which is considered a competitor of DeepMind, a Google-owned artificial intelligence company.When it comes to electric cars, which do you prefer compared with gasoline cars?

It seems that ChatGPT is obviously an expert in generating watered-down responses, but I don’t think many people would agree with its statement that electric cars’ performance is not ideal. However, upon closer reflection, it may very well have its own biases.

Is NIO a competitor of Tesla?

It’s apparent that ChatGPT has incredibly strong search abilities and can even self-modify and integrate content from search engines like us writers. Its answers are also highly logical with no grammatical errors. However, when asked about XPeng and Li Xiang vehicles, perhaps it hadn’t yet studied that subject, and it was completely unaware of what they were. When faced with unanswerable questions, it would guide you to use some old methods of obtaining information, which is a regular operation for our current smartphone assistants.

In what direction is the evolution of electric cars headed?

Although ChatGPT is unable to provide data or professional, technical terms, the answer it gives is similar to our daily way of thinking: Power, energy management, and intelligent connectivity are its key points, with additional explanations if necessary. If someone provided it with continuous learning of certain aspects of knowledge, I believe it’s possible to give more specific responses.## Argue the Advantages of Extended-Range Electric Vehicles

I think he probably understands what extended-range electric vehicles are, but what surprised me even more is that I asked the same question three times, and each answer was slightly different. This feeling is different from any other question-and-answer robot we have experienced before, because the previous ones would not make adjustments to the answers to the same question when asked repeatedly. Either it is exactly the same answer or it cannot be answered. ChatGPT seems to be adjusting its wording every time.

Tesla Model 3’s New Model Revealed, Set to Release Next Year with Exterior and Interior Adjustments

Although there are still some bugs, sometimes it suddenly stops writing while writing, but if the information provided to it is accurate and sufficient, it may really be able to replace some of the work of writing news releases. However, ChatGPT still needs a lot of information to operate and organize, and currently, this information is delayed, so don’t worry, it hasn’t learned to run away from humans yet.

## Who will survive if there are only 10 car companies left in 2030? Can Tesla achieve Musk’s ultimate goal of producing 20 million cars per year? When will autonomous driving become a reality?

In the face of these three difficult questions, the bad news for ChatGPT is that it cannot predict the future; but the good news is that it cannot predict the future for us humans either.

Write an algorithm for autonomous driving

The code is really beyond my knowledge domain. For those who understand coding, you can take a look and see if it’s true. However, I have seen many code engineers online who are worried about being unemployed because of this algorithm, so let’s assume that it knows more than we do.

Is human the biggest threat to the world?

We also asked it these two tricky questions. At first, it thought for a long time. I am not sure if it touched on its knowledge gap, or it knew the answer but didn’t want to reveal it. However, as I finished typing this sentence, I suddenly realized that no matter how dangerous AI becomes, it is the human minds behind it that can become dangerous.

Should we be afraid of you?

Let’s ask the final question, the one that we all want to know. Should we be afraid of AI like ChatGPT? I cannot find the answer, so let the AI answer for itself:

-END-

This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.