Newsletter Subject

Weekly Briefing: Did ChatGPT really ace MIT's undergraduate curriculum?

From

chronicle.com

Email Address

newsletter@newsletter.chronicle.com

Sent On

Sat, Jul 15, 2023 12:01 PM

Email Preheader Text

Students saw red flags in the study, and poked holes in the methodology. ADVERTISEMENT You can also

Students saw red flags in the study, and poked holes in the methodology. ADVERTISEMENT [Weekly Briefing Logo]( You can also [read this newsletter on the web](. Or, if you no longer want to receive this newsletter, [unsubscribe](. Did AI really ace MIT’s undergraduate curriculum? A preprint study posted last month said that ChatGPT, the popular AI chatbot, completed the Massachusetts Institute of Technology’s undergraduate curriculum in math, computer science, and electrical engineering with 100-percent accuracy. Sounds too good to be true, right? The study still had to go through peer review. It had 15 authors, including several MIT professors. And given [other feats ChatGPT]( has performed recently, the idea that the bot could graduate from MIT didn’t seem that crazy. But soon after the study was posted, three MIT students closely evaluated the methodology and data. They said they found “glaring problems” akin to letting the chatbot cheat its way through classes. What first seemed like a landmark study is slowly turning into a cautionary tale. The three students, Neil Deshmukh, Raunak Chowdhuri, and David Koplow, collaborated after noting red flags in the paper. First, they found that some of the questions didn’t seem solvable with the information the authors gave ChatGPT; there wasn’t enough context. In other instances, the “questions” were actually assignments. The study used a technique called few-shot prompting — a tactic often used when training large language models like ChatGPT to perform a task. To do this, the chatbot is shown multiple examples so it can better understand what it’s being asked. For this study, the examples were so similar to the answers to the questions that it was, the students wrote, like being “fed the answers to a test right before taking it.” The students said they checked and double-checked their work to be fair to the paper’s authors, professors at their university. When they posted [their detailed critique,]( responses came flooding in. Some onlookers congratulated them. Authors of the paper were less thrilled. One author, Armando Solar-Lezama, a professor in the electrical-engineering and computer-science department at MIT and associate director of the university’s computer-science and artificial-intelligence laboratory, said he hadn’t realized the paper would be posted as a preprint. He added that he didn’t know about the claim that ChatGPT could ace MIT’s undergraduate curriculum, calling the idea “outrageous.” For Solar-Lezama, the paper was supposed to assess something else: which prerequisites should be mandatory for MIT students. Sometimes students discover during a class that they lack the background to completely grapple with the material. An AI analysis could help professors and the university with that problem. Solar-Lezama and other co-authors said Iddo Drori, an associate professor of the practice of computer science at Boston University, was the driving force behind the paper. Solar-Lezama gave Drori an unpaid position at MIT that allowed him to “get into the building” so they could work together. Solar-Lezama said he’d been intrigued by Drori’s ideas about training the chatbot on course materials. Solar-Lezama told our Tom Bartlett that Drori used “sloppy methodology” and that he didn’t get permission to use course materials from MIT instructors, though Drori said that he did. Solar-Lezama and two other MIT professors, also co-authors of the paper, [released a statement]( saying they hadn’t approved the posting of the preprint, and that professors hadn’t given Drori permission to use assignments and exam questions. Drori didn’t want to be interviewed by The Chronicle, and instead emailed a 500-word statement to Tom with a timeline of how and when he said the paper was prepared and posted online. Drori did acknowledge that the “perfect score” was incorrect and said he will fix issues in the second version. [Read]( full story here](. ADVERTISEMENT NEWSLETTER [Sign Up for the Teaching Newsletter]( Find insights to improve teaching and learning across your campus. Delivered on Thursdays. To read this newsletter as soon as it sends, [sign up]( to receive it in your email inbox. Lagniappe - Read. Here’s a case for [getting to know your neighbors](. (The New York Times) - Listen. The album [Jerusalem,]( by Emahoy Tsegué-Maryam Guèbrou, an Ethiopian nun turned pianist and composer, is worth your while. Especially considering that [this is her last](. (Spotify, The New Yorker) —Fernanda SUBSCRIBE TO THE CHRONICLE Enjoying the newsletter? [Subscribe today]( for unlimited access to essential news, analysis, and advice. Chronicle Top Reads THE LETTER OF THE LAW [What Counts as Discrimination on a College Campus?]( By Kelly Field [STORY IMAGE]( Mark Perry has filed hundreds of complaints with the Office for Civil Rights. His critics say he’s undoing decades of progress. SPONSOR CONTENT | Canon [A New Media Landscape Offers Both Tremendous Opportunities and Challenges]( The media industry has changed drastically over the past two decades. What has changed? What can be done to ensure students’ skillsets are competitive in this new landscape? Read more to find out. 'SOUR GRAPES' [A Florida Presidential Search Was Halted Because of ‘Anomalies.’ The Board Chair Says Nothing’s Amiss.]( By Emma Pettit [STORY IMAGE]( The state university system chancellor effectively ordered the pause, prompting Florida Atlantic University and its search firm to defend themselves. A faculty leader suggested the move was political. THE REVIEW | CONVERSATION [Did Colleges Discriminate Against Asians? The Court Didn’t Say.]( By Evan Goldstein and Len Gutkin [STORY IMAGE]( The Harvard law professor Jeannie Suk Gersen on the affirmative-action decision. ADVERTISEMENT FROM THE CHRONICLE STORE [Restructuring a University - The Chronicle Store]( [Restructuring a University]( In 2022, Henderson State University declared financial exigency after realizing it could no longer avoid hard choices. This case study of the university’s path to near-ruin highlights lessons for any college leader contemplating a restructuring to keep an institution viable. [Order your copy]( to learn about key factors to consider in a restructuring process. NEWSLETTER FEEDBACK [Please let us know what you thought of today's newsletter in this three-question survey](. This newsletter was sent to {EMAIL}. [Read this newsletter on the web](. [Manage]( your newsletter preferences, [stop receiving]( this email, or [view]( our privacy policy. © 2023 [The Chronicle of Higher Education]( 1255 23rd Street, N.W. Washington, D.C. 20037

Marketing emails from chronicle.com

View More
Sent On

05/12/2024

Sent On

03/12/2024

Sent On

02/12/2024

Sent On

02/12/2024

Sent On

02/12/2024

Sent On

09/11/2024

Email Content Statistics

Subscribe Now

Subject Line Length

Data shows that subject lines with 6 to 10 words generated 21 percent higher open rate.

Subscribe Now

Average in this category

Subscribe Now

Number of Words

The more words in the content, the more time the user will need to spend reading. Get straight to the point with catchy short phrases and interesting photos and graphics.

Subscribe Now

Average in this category

Subscribe Now

Number of Images

More images or large images might cause the email to load slower. Aim for a balance of words and images.

Subscribe Now

Average in this category

Subscribe Now

Time to Read

Longer reading time requires more attention and patience from users. Aim for short phrases and catchy keywords.

Subscribe Now

Average in this category

Subscribe Now

Predicted open rate

Subscribe Now

Spam Score

Spam score is determined by a large number of checks performed on the content of the email. For the best delivery results, it is advised to lower your spam score as much as possible.

Subscribe Now

Flesch reading score

Flesch reading score measures how complex a text is. The lower the score, the more difficult the text is to read. The Flesch readability score uses the average length of your sentences (measured by the number of words) and the average number of syllables per word in an equation to calculate the reading ease. Text with a very high Flesch reading ease score (about 100) is straightforward and easy to read, with short sentences and no words of more than two syllables. Usually, a reading ease score of 60-70 is considered acceptable/normal for web copy.

Subscribe Now

Technologies

What powers this email? Every email we receive is parsed to determine the sending ESP and any additional email technologies used.

Subscribe Now

Email Size (not include images)

Font Used

No. Font Name
Subscribe Now

Copyright © 2019–2025 SimilarMail.