OpenAI releases ChatGPT classroom usage guide for the first time, Harvard and Yale also start using AI as teachers

The back-to-school season is here, and many of the post-2000s who entered the teaching profession have also ended their first summer vacation as teachers.

Students agree with many teachers born after 2000 in not wanting to start school.

From another perspective, those post-2000 teachers who don’t want to go to work may be the same group of students who once didn’t want to go to school.

Just like adults who take a long vacation, they will also have anxiety about going to work, and teachers are no exception. In response to teachers’ “school anxiety”, OpenAI has given a “good remedy”.

How teachers should use ChatGPT

On August 31, OpenAI released for the first time a guide for teachers on the use of ChatGPT, including how ChatGPT works, suggested prompt words, explanations of limitations, and the effectiveness and bias of AI detectors.

It is worth noting that this is also OpenAI’s first official industry-specific usage guide. OpenAI also shared some cases of teachers using ChatGPT to assist students in learning on its official website.

Fran Bellas, a professor at the University of Coruña in Spain, recommends that teachers use ChatGPT to help develop classroom tests, exams and lesson plans, obtain new teaching ideas, and ensure that questions are appropriate for students' levels.

For students, if you encounter difficulties in finding an internship in your junior year, you can also use ChatGPT to simulate an interviewer. Dr. Helen Crompton of Old Dominion University encourages students to improve their interview skills and understanding in this way.

More specifically, if you want to improve work efficiency, you might try the prompts (Prompts) officially shared by OpenAI.

For example, with the help of prompt words, you can ask ChatGPT to help you plan lessons and create easy-to-understand examples to better explain knowledge points to students.

Or let students play the role of teachers, point out errors in ChatGPT and provide corrections and guidance to improve students' understanding of knowledge points.

Of course, good education is not about "telling" but "guiding". ChatGPT can also guide students to learn new knowledge points independently, cultivate their critical thinking and problem-solving skills, and enable them to better deal with complex knowledge areas.

question, understand, become

ChatGPT became an overnight sensation at the end of last year, with countless people waving the flag and shouting about generative AI, as if the opportunity to reshape thousands of industries was right in front of them.

Even though it is the most prestigious university in the world, Harvard University’s concept has undergone several changes in the face of the wave of generative AI coming towards the education world. From questioning, understanding to becoming, Harvard University can be said to be "Really fragrant warning".

In January this year, Anne Harrington, acting dean of education at Harvard University, stated in an internal email that Harvard’s honor code “prohibits students from treating as their own work work they have not written, coded, or created.” The implication is to warn students not to use generative expressions easily. AI completes the work.

▲ Picture from: The Harvard Crimson

Nowadays, perhaps because of the realization that "blocking is worse than sparseness" of AI's intrusion into the classroom, or the realization of the promising prospects of the wave of AI technology, Harvard University's "iceberg" attitude towards generative AI is gradually melting.

In July, Harvard University released a preliminary guide for the use of generative AI across the school. The guide clearly emphasizes the protection of confidential data, attention to the authenticity of generated content, compliance with current academic integrity policies, and vigilance against artificial intelligence phishing. Content.

In order to further protect students’ confidential data, Harvard University’s School of Information Technology is also jointly developing a tool called “AI Sandbox” with a third-party artificial intelligence company for use by Harvard University-affiliated institutions.

Jason A. Newton, a spokesman for Harvard University, said: "This tool provides an isolated and safe environment to effectively avoid security and privacy risks when using generative AI and ensure that the input data will not be used for training. Model."

Recently, the Office of Undergraduate Education in Harvard University's College of Arts and Sciences released a guide on using generative AI such as ChatGPT in the classroom, aiming to better help teachers understand how AI works and its potential applications in teaching.

Instead of enforcing it, the guide recommends that teachers can adopt three distinct attitudes when using AI in their courses: a completely restrictive attitude, a fully encouraging attitude, or a neutral attitude.

Christopher W. Stubbs, dean of the College of Arts and Sciences, said: "We ask faculty to understand how generative AI will impact the curriculum…It is important that they (faculty) clearly communicate their selected course attitudes to students."

Twenty-four top universities, including Oxford and Cambridge, which had banned the use of ChatGPT in the past few months, have also "slapped" them in the face and signed new guidelines to allow students and faculty to use ChatGPT reasonably and provide guidance on how to use it.

"These policies make it clear to students and staff where the use of generative AI is inappropriate, and are designed to support them in making informed decisions and empower them to use these tools appropriately and, where necessary, acknowledge their use," the new guidance says. .”

Generative AI teaches you how to write papers

The change in the attitude of top universities is more like a small slice of generative AI’s “invasion” of the education field.

In foreign countries, many top universities often require an application to attend. How to stand out in the "Application Questionnaire" (the question section on the application form that needs to be filled out when applying to university) has become the ultimate BOSS that every high school student needs to overcome on their way to success. .

New York Times reporter Natasha Singer used generative AI tools to test application questions from several top universities. We also drew on similar ideas to see how the feedback capabilities of generative AI are and whether they can help high school students get their favorite college offers. ?

One of Harvard's short-answer questions asks applicants to describe: "Three things your roommate would most like to know about you."

▲ChatGPT’s answer

A short-answer question from Yale University is: "If you could teach any college course, write a book, or create original artwork of any kind, what would you do?"

▲ Claude’s answer.

A short-answer question from Princeton University is "Which song represents the soundtrack of your life at this moment?"

▲Perplexity’s answer.

Natasha Singer later also interviewed staff at some universities’ admissions offices. Staff at Yale University believe that using tools such as ChatGPT to write college essays is essentially a form of plagiarism, and that applicants who submit essays generated by chatbots will violate the university’s admissions policies. .

In addition, they also believe that the essay for college application "means introspection and reflection" and "outsourcing" personal thinking to AI is undesirable behavior.

However, Jenny Frederick, vice provost of Yale University, recently said: "Yale University has never considered banning the use of ChatGPT."

In Frederick's view, artificial intelligence will be integrated into all walks of life in different ways, and students need to be prepared for this. "We really need to prepare them."

Virginia Tech's Espinosa and other admissions staff also told Natasha Singer that for many high school students with limited or no experience applying to college, tools such as ChatGPT can reduce the "experience" gap and benefit social equity.

After generative AI enters the classroom, it becomes more important than combating “cheating”

The views on generative AI may be different. If students are determined to "open cheats", teachers may be helpless.

Maya Bodnick, a politics major at Harvard University, invited seven professors and teaching assistants to grade GPT-4 papers written based on class prompts. In order to test the real evaluation, Maya Bodnick lied that the authors of these papers were either Maya Bodnick or GPT-4. In the end, Maya Bodnick was pleasantly surprised and received two A's, one A-, one B, and one B-. A C.

This shows that the paper ability of generative AI is even comparable to the passing level of most liberal arts majors. When students hold the artifact of generative AI, it is obviously not a problem for teachers who do not have "sharp eyes" to detect students' "plug-ins". Simple things.

"Although there are multiple tools claiming to have varying degrees of success in detecting generative AI, these methods are unreliable," College of Arts and Sciences officials said.

▲Picture from: MIT Technology Review

In previous articles, we also reported on the problem of “counterfeiting” of GPT detectors. OpenAI, which is good at using "magic to defeat magic", did not stage a cat-and-mouse game after launching its text detector (AI Text Classifier). Since the recognition accuracy rate was only 26%, it was released just a few months later. Set up this tool.

What's more, generative AI tools are not necessarily entirely bad for students.

For example, Khan Migo, a chatbot from the online education nonprofit Khan Academy, kindly informs users before suggesting ideas for students’ college admissions essays: “A college essay is not a place to share a list of your accomplishments, but a place to showcase your uniqueness. Opportunity for personality and perspective.”

In addition, rather than directly providing answers, Khanmigo focuses more on helping students develop new ideas in their college admissions essays, guiding students to gradually improve the content of their essays with step-by-step instructions.

Perhaps as Salman Khan, the founder of Khan Academy, envisions:

AI will enable the greatest positive change in education ever, providing every student on the planet with an incredible personal tutor and every teacher with an extraordinary teaching assistant.

#Welcome to follow Aifaner's official WeChat public account: Aifaner (WeChat ID: ifanr), more exciting content will be presented to you as soon as possible.

Ai Faner | Original Link · View Comments · Sina Weibo