'ChatGPT was blocked by schools and major companies, causing social panic'

In December, when ChatGPT was becoming popular, many scholars and researchers warned that ChatGPT was likely to "kill" university papers. Recently, the New York City Department of Education officially issued a ban order on ChatGPT.

Coincidentally, ChatGPT has also been blocked by many major Internet companies. Many coders who have become accustomed to consulting ChatGPT about coding issues burst into tears...

Big manufacturers blocked ChatGPT, and coders cried

Recently, a A coder posted online that in his large factory, ChatGPT was disabled on the office wifi. This immediately broke him down.

The blogger said that he is a junior software engineer, and in the team, the seniors are always busy and have no time to explain the problem to him .

ChatGPT is his savior whenever he gets stuck at work.

But yesterday, when he sat down in the office, he suddenly found that he could not access ChatGPT! He broke through the defense instantly.

Under his post, enthusiastic netizens offered their suggestions.

For example, buy a cheap VPS and domain name and set up a reverse proxy to the ChatGPT official website; enable DoH and ECH in the browser; use an online proxy; or use the Email the output to yourself and so on.

Of course, some netizens agreed with the company's operation: "It makes sense for your company to ban ChatGPT. Think about it carefully. If you The code that solves the company's problems can be easily obtained by competitors, what will be the consequences?"

"Do not submit requests to ChatGPT, or from ChatGPT Copy and paste the code on ChatGPT. Legally, the company or client's code does not belong to you. And you do not have the copyright to the code on ChatGPT. If you do this, you will face lawsuits and be fired."

Indeed, some major manufacturers have sent emails to employees saying: "ChatGPT is not unusable, but if you upload code to it, you will face the risk of being fired, because our code is very sensitive."

However, some netizens gave a heart-wrenching reply: "If a junior developer cannot complete his work without ChatGPT, then he is really valuable to the company. It’s dispensable.”

New York issued a ban, prohibiting all teachers and students

On January 3, New York City, which has the largest public school system in the United States, The "ChatGPT ban" was officially promulgated.

In other words, neither teachers nor students can use ChatGPT on the networks and devices of New York City public schools.

Recently, school districts in Los Angeles and Baltimore have jumped on the bandwagon.

New York City Department of Education spokesperson Jenna Lyle said that while ChatGPT can provide quick and easy answers to questions, it Failure to develop students’ critical thinking and problem-solving skills, which are critical to academic and lifelong success.

Of course, if the school’s purpose is to study the technology behind ChatGPT, it can still apply for access.

For now, it remains to be seen what impact ChatGPT will have on the risk of cheating on essays, as students will still have access to their own devices and networks off campus.

The attitude of Cobb County in Georgia is not only not so tough, but is even quite interested in adopting ChatGPT.

A spokesperson for the district said, "In the hands of highly qualified Cobb County teachers, ChatGPT can be a valuable tool that supplements students' understanding of academic content and can help students develop Critical thinking skills."

In this regard, Adam Stevens, who teaches at Brooklyn Tech, an elite high school in New York, also believes that banning ChatGPT will have a negative impact.

After all, 15 or 20 years ago, when students could "find answers online," these people were targeting Google in the same way.

Stevens believes that the best way to develop students' critical writing skills is to "invite them to explore things worth knowing" and move away from formulaic writing instruction based on strict grading standards.

"Of course, if you value the grade at the end of the semester more, then ChatGPT is indeed a threat."

Similarly, Sandy, Oregon A high school English teacher in the city also had a similar view.

Teacher Shields has 25 years of teaching experience. She believes that in fact, there is no need to "talk about ChatGPT." This AI can be used to help students generate ideas, organize ideas, and even Write the entire paper.

In her opinion, teachers should accept that AI is a tool that students can use, just like every teacher teaches students how to use search engines.

Many students struggle with having nothing to write. At this time, teachers can use ChatGPT to show students how to generate prompts to stimulate their writing.

Teacher Shields once asked ChatGPT to write about the same topic in the style of a 7th grader, a 10th grader, and a college student, and the degree of syntax difference between each level was astounding. Therefore, it is entirely possible to use this AI-generated feedback to show students how to improve their writing.

ICML issued a clarification: It is mainly to prevent plagiarism and does not affect the polishing of articles

Two days ago, the International Machine Learning Conference ICML 2023 suddenly announced that the use of large-scale language models is prohibited. (e.g. ChatGPT) generates the content of a paper, unless the text is part of an experimental analysis.

Once the request was made, it aroused the dissatisfaction of many researchers: spelling check and text prediction are based on language models, why are we not allowed to use them?

In this regard, Turing Award winner Yann LeCun commented: "Large language models cannot be used, which means that they can be replaced by medium and small language models. Really?"

Then I didn’t forget the yin and yang sentence: "Next year’s ICML code of ethics should become: researchers must complete various calculations with bare hands and cannot use computers; all types of graphics must be done with the help of computers. Hand-drawn with pen and paper; starting today, the official language of ICML becomes Swedish (Sweden:?)."

Because the discussion caused is too It was so drastic that ICML had to issue a long article explaining this rule. In summary, there are three main points:

• The policy prohibits text that is completely "generated" by LLM, but this does not affect the author's use of LLM to edit or polish his own text.
• The policy is mainly to prevent potential problems that may arise when using LLM, such as plagiarism.
• Currently, the policy only applies to ICML2023. This policy may change at future meetings as we gain a greater understanding of LLMs and their impact on scientific publishing.

## Announcement link: https://icml.cc/Conferences/2023/llm-policy

Large-scale language models, full of controversy

In the past few years, large-scale language models (LLM) have developed very rapidly. Lately, this progress has not slowed down, it has even accelerated.

As you can see, LLMs released in the past few months, such as OpenAI’s ChatGPT, have been able to generate text fragments that are difficult to distinguish from humans. There is no doubt that this is an exciting advance in natural language processing and generation.

But this rapid progress is often accompanied by unintended consequences and unanswerable questions. For example, are the text and images generated by the model novel, or are they derivatives of existing work?

In addition, who should own the text snippets, images, and other content produced by these models: the user who generated the model, the developer who trained the model, or the person who created the training Content creator of data?

To be sure, these questions will be answered over time as large-scale generative models become more widely used, but not yet. Especially less than 2 months have passed since the release of ChatGPT (the end of November 2022).

For ICML, they do not have enough time to observe and investigate the impact ChatGPT will have on the paper review and publication process.

Considering the strategies adopted for this new technology will directly affect the careers of researchers. Therefore, ICML said that it can only be more cautious and even a little conservative.

There are mistakes and omissions from time to time, but coders still love it

In fact, although ChatGPT is often ridiculed for being "stupid", in the more than a month since its birth, The working habits of many programmers have indeed been changed by it.

After ChatGPT was unveiled in early December, coders immediately became excited when they discovered that it could be used to generate code.

ChatGPT can not only answer questions in any field, but the answers generated have a high probability of being very accurate and detailed.

For example, when you ask "How to iterate over an keys and values ​​array in Javascript", it generates a perfect answer.

The icing on the cake is that ChatGPT can also tell you the best solution to this common problem in JavaScript through a supporting example and explanation:

You can even use it to correct code errors.

But then, coders immediately discovered that the code generated by ChatGPT actually had many problems.

ChatGPT will be "stupid". The reason behind it is actually very simple - the output of AI is probabilistic. ChatGPT does not have any internal records that mark right or wrong. It is actually a statistical model. In this model, it is a combination of languages ​​in different contexts.

The basis of context is the overall data set used for GPT-3 training, additional context from ChatGPT’s RLHF training, prompts and previous conversations, and what will be available soon feedback.

ChatGPT’s answers are essentially probabilistic results collected from the corpus of Internet data that makes up GPT-3. Within 10 seconds, ChatGPT makes its best guess at the outcome, which is likely to be correct.

Of course, there is also the possibility of error.

This is why, when netizens used the code generated by ChatGPT to flood Stack Overflow with incorrect answers, the latter issued an official ban order overnight.

Although ChatGPT’s ability to write code has been criticized by many people, in fact, it can still be used.

In fact, the most suitable positioning of ChatGPT should be a coding assistance tool. According to OpenAI, the publisher of ChatGPT, one of the important uses of ChatGPT is to help programmers check code.

For developers, it is just one tool in an increasingly rich palette of tools. And, like any self-learning model, ChatGPT will continue to learn, self-correct and improve over time.

Many times, instead of waiting for days on Stack OverFlow for a relatively simple question (how to trigger the rebuild of HookConsumerWidget in Flutter), it is better to ask the question directly to ChatGPT.

Because on ChatGPT, developers have a high probability of getting a starting point that can be used to solve problems——

The above is the detailed content of 'ChatGPT was blocked by schools and major companies, causing social panic'. For more information, please follow other related articles on the PHP Chinese website!