Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizen Coje_He for the clue delivery!

The birth of ChatGPT has rewritten the rules of the game in many fields. Now, in many places, schools and big factories have issued a ChatGPT ban.

In December, when ChatGPT became so popular, many scholars and researchers warned that ChatGPT was likely to "kill" college papers. Recently, the New York City Department of Education officially banned ChatGPT.

Coincidentally, in many big Internet companies, ChatGPT has also been banned. Many code farmers, who are used to asking ChatGPT about code problems, burst into tears in place.

Programmers cried recently after the big factory banned ChatGPT. One programmer posted online that in his big factory, the office wifi banned ChatGPT. It just freaked him out.

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.

Whenever he gets stuck at work, ChatGPT is his savior.

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

Under his post, enthusiastic netizens offered suggestions one after another.

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 to use an online proxy; or email the output to yourself on your mobile phone, and so on.

Of course, some netizens agree with this operation of the company: "it makes sense for you to disable ChatGPT. Think about what will happen if your code to solve the company's problems can be easily obtained by competitors." "

"do not copy and paste code to ChatGPT or from ChatGPT. Legally, the company or customer code does not belong to you. And you don't have the copyright of the code on ChatGPT. If you do this, you will face litigation and dismissal. "

Indeed, a big company has sent an email to its employees saying, "ChatGPT is not disabled, but if you upload code to it, you will be at risk of being fired because our code is very sensitive." "

However, some netizens gave a heartbreaking reply: "if a junior developer cannot finish his work without ChatGPT, then he is really dispensable to the company." "

New York issued a ban on all teachers and students on January 3, New York City, which has the largest public school system in the United States, formally issued a "ChatGPT ban."

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

Recently, school districts in Los Angeles and Baltimore have joined the ranks.

Jenna Lyle, a spokesman for the New York City Department of Education, said that while ChatGPT can provide quick and simple answers to questions, it does not develop students' critical thinking and problem-solving skills, which are essential for academic and lifelong success.

Of course, if the purpose of the school is to study the technology behind ChatGPT, you can still apply for access.

At present, it remains to be seen what impact ChatGPT will have on the risk of paper cheating, as students can still access their own devices and networks off-campus.

Cobb County, Georgia, is not only less tough, but even interested in adopting ChatGPT.

A spokesman for the school district said, "in the hands of high-quality Cobb County teachers, ChatGPT can be a valuable tool to supplement students' understanding of academic content and help students develop critical thinking skills. "

In response, 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, that's what these people did to Google 15 or 20 years ago, when students could "find answers online."

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

"of course, if you value end-of-term grades more, then ChatGPT is a real threat. "

Similarly, a high school English teacher in Sandy, Oregon, has a similar view.

Shields, who has been teaching for 25 years, believes that there is no need to "talk about ChatGPT color change". The AI can be used to help students generate ideas, organize ideas, and even write an entire paper.

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

Many students suffer from having nothing to write, so teachers can use ChatGPT to show students how to generate hints to stimulate their writing.

Teacher Shields once asked ChatGPT to write articles on the same topic in the style of seventh graders, 10th graders and college students, and the syntactic differences between each level were shocking. Therefore, the feedback generated by AI can be used to show students how to improve their writing.

ICML post clarification: mainly to prevent plagiarism, do not affect retouching articles two days ago, the International Machine Learning Conference ICML 2023 suddenly announced that it is forbidden to use large-scale language models (such as ChatGPT) to generate the content of papers, unless these texts are part of experimental analysis.

As soon as the request was made, it aroused the dissatisfaction of many researchers: spell checking and text prediction are based on language models, so why not let us use them?

In response, Yann LeCun, a Turing Award winner, commented: "large language models cannot be used, meaning medium and small language models are fine?" "

Then do not forget Yin and Yang: "next year's ICML code of ethics should be: researchers must do all kinds of calculations with their bare hands, not with the help of computers; all kinds of graphics must be drawn by pen and paper; from today, the official language of ICML will become Swedish (Sweden:? ). "

The discussion was so heated that ICML had to send a long message to explain the rule. To sum up, there are three main points:

The policy forbids text that is "generated" entirely by LLM, but this does not affect authors' use of LLM to edit or polish their own text.

The policy is mainly to prevent potential problems that may arise from using LLM, such as plagiarism.

Currently, the policy applies only to ICML2023. As we learn more about LLM and its impact on science publishing, this policy is likely to change in future meetings.

Bulletin link: https://icml.cc/ Conferences / 2023 / llm-policy large-scale language model, controversial in the past few years, large-scale language model (LLM) has developed rapidly. Recently, this progress has not only not slowed down, but even faster.

As you can see, LLM released in the past few months, such as OpenAI's ChatGPT, has been able to generate text fragments that are indistinguishable from humans. There is no doubt that this is an exciting development in natural language processing and generation.

But this rapid progress is often accompanied by unexpected consequences and unanswered 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 clips, images, and other content generated by these models: the user who generated the model, the developer who trained the model, or the content creator who produced the training data?

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

As for ICML, they do not have enough time to observe and investigate how ChatGPT will affect the review and publication process of papers.

Considering the strategies used for this new technology, it will have a direct impact on the careers of researchers. So ICML says he can only be more cautious, even a little conservative.

There are mistakes and omissions from time to time, but programmers still love it. In fact, although ChatGPT is often laughed at for being "stupid", it has changed the work habits of many programmers since it was born more than a month ago.

After the debut of ChatGPT in early December, the code farmers found that they could use it to generate code, and immediately became agitated.

ChatGPT can not only answer questions in any field, but also generate answers with 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 produces the perfect answer.

To add to the icing on the cake, 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 programmers immediately discovered that there were a lot of problems with the code generated by ChatGPT.

The reason behind ChatGPT being "stupid" is actually very simple-the output of AI is probabilistic. ChatGPT does not have any internal records marked right or wrong, it is actually a statistical model, which is a combination of languages in different contexts.

The context is based on the overall data set used for GPT-3 training, additional context from ChatGPT's RLHF training, prompt and previous conversations, and feedback soon.

ChatGPT's answer is essentially the probabilistic results collected from the Internet data corpus that makes up GPT-3. Within 10 seconds, ChatGPT made the best guess about the result, which is probably correct.

Of course, there is also a probability of error.

That's why, when netizens flooded Stack Overflow crazily with ChatGPT-generated code and the wrong answers were everywhere, the latter issued an official ban overnight.

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

In fact, the most suitable location for ChatGPT should be coding aids. According to OpenAI, the publisher of ChatGPT, one of the important uses of ChatGPT is to help programmers check the code.

For developers, it is a tool in an increasingly rich toolboard. And, just like any self-study model, ChatGPT will continue to learn, correct and improve itself over time.

In many cases, instead of waiting a few days on Stack OverFlow for a relatively simple question (how to trigger the reconstruction of HookConsumerWidget in Flutter), ask ChatGPT directly.

Because on ChatGPT, developers have a good chance of getting a starting point that can be used to solve problems--

Reference:

Https://www.businessinsider.com/nyc-schools-ban-chatgpt-cheating-concern-grows-2023-1

Https://www.theregister.com/2023/01/06/ai_conference_nyc_ban/?td=rt-3a

Https://www.reddit.com/r/ChatGPT/comments/104rp2j/my_company_blocked_chatgpt/

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.