ChatGPT’s, an artificial intelligence bot programmed to simulate human conversation through text, responses to respective prompts were graded by three Long Beach City College professors and it was capable of producing passable submissions for every professor.
The bot was designed by Open AI, an American artificial intelligence research laboratory, to communicate with users in a natural language format.
The basis of what ChatGPT is used for could be revolutionary for humans, but it has the potential to be used in a different way than intended.
Cheating.
“AI is imitating us! Why are we bowing down to it,” asked Dr. Margaret Shannon, an English professor who graded ChatGPT.
The other professors who graded were Dr. Gilbert Estrada, a history and ethnic studies professor, and Dr. Franklin Perez, an ethnic studies professor.
All of them said it would be a passable assignment in their respective courses with differing grades.
The bot scored as high as an A on one of Perez’s assignments, but the average was a C or barely passable. Perez’s prompt included what articles to cite from and only then was the bot able to properly cite information, but not direct quotes.
Each professor criticized the bot in not being able to properly provide sources for their work, which docked the AIs grade the most.
The professors pointed out that they use specific sources for their courses which might act as a defense to AI cheating.
Another defense against cheating was the professors assigning a variety of assignments from presentations, essays and tests.
The professors also brought up that one essay from a student that is AI generated out of the 200 students they teach can easily get by and pass for a student written essay.
Though one AI generated might slip through, the professors said if multiple AI generated essays were submitted it would be an obvious red flag.
After inputting multiple assignments, professors realized the essays generated were “template-like.” The program spit out essentially the same essay using the different language used in the prompt. Essay structures were almost identical if asked for an argumentative essay or admissions essay.
The advancement of technology comes with the risk of it being exploited and used for personal gain.
Software like ChatGPT is not going unnoticed and counter measures are being developed to mitigate some of the potential issues.
Programs like GPTZero aim to distinguish human text from AI generated text.
ChatGPT is capable of passing AP exams, the bar exam, writing plays or helping with content creation. ChatGPT not only is capable of doing these tasks, but it does it well.
The bot scored on the University of Minnesota bar exam within the top 10%.
ChatGPT works by using a text database called a corpus. The corpus contains millions of words and sentences from literature, web pages, news articles and social media. It uses deep learning to analyze and recognize speech patterns to formulate its own ideas.
Deep learning is a method of artificial intelligence that teaches computers to process data in the way a human does.
The developers of ChatGPT are constantly improving the AI so that it can better assist humans.
GPT-4 is the newest version of ChatGPT and this version is now able to read images.
This means if a person uploaded a picture of eggs, flour, butter and milk and asked the AI: “What can I make with this?”, the AI would be able to; “read” your image, realize the things in the picture and give the user options on what can possibly be made out of the ingredients provided.
ChatGPT has decent knowledge on how the world works already and with time we can only imagine what it will continue to learn.
There’s no way of predicting the future capabilities of AI. Perhaps it will completely replace humans or force them out of even the safest jobs.
Is AI capable of that? No…
Not yet.
| Professor Grades |
| Margaret Shannon, English Professor Grade: “Courtesy” C / Not Admitted Shannon was critical of the AI response to an essay prompt she came up with. She plugged in a number of essays, from argumentative essays to admissions essays. “ChatGPT writes like a mediocre mind,” Shannon said. “It had some little nuggets buried, if you gave it a more specific prompt. This is a perfect solution for those who don’t value education and want to leave here unchanged. Why are we trusting this thing for a refined answer when people don’t know how to ask good questions?” Shannon ultimately gave the AI response a C because it conveyed some sort of deeper thought to the prompts she put it in. She also explained if someone submitted a ChatGPT admissions essay, the user would not get in. | Gilbert Estrada, History and Ethnic Studies Professor Grade: Not passing / Passable Estrada was confident his methods of using specific prompts and sources can counter the power of ChatGPT. He input an Aztec informational essay and the AI gathered information he does not cover in class which was a giveaway something was fishy. “I have specific sources so if I were to be turned in, it’s not what I’m looking for,” Estrada said. “[It’s] decent writing, it can easily fool me if there were some data points. I still think I can catch it, but again, ask me in a year or two and my confidence will waiver.” The second essay Estrada input was a simple information gathering essay about a Tenochtitlan. Most of the information was accurate to Estrada’s course, but some facts stood out that he doesn’t cover. Overall, he stated the second essay would have received a passing grade. | Franklin Perez, Ethnic Studies Professor Grade: A / C Perez was shocked to see the AI was able to link concepts to each other in his input of an essay on race. ChatGPT was able to link race, racialization, and colorism and provide examples to support its thesis. The second essay was an argumentative essay asking the AI to cite a film and take a stance whether the war on drugs is a slow motion holocaust. The AI was not able to cite the film and sources provided, but still could pass with a C. “I have no way of saying a student or ChatGPT did this, but there’s no clear indication that the AI watched the film,” Perez said. “People used to cheat by getting access to standardized tests before they were given and it started a movement towards essays. If SLOs dictate the effectiveness of the professor, but people can effectively cheat, what do we do now?” |