This Stanford computer science professor went to written exams 2 years ago because of AI. He says his students insisted on it
Stanford University Professor of Computer Sciences Jure leskovec It is not foreign to rapid technological change. For about thirty years and for the second decade, a machine learning researcher, as well as the founding partner of Kumo, 37 million dollars of financing collected so far.
But two years ago, when the latest wave of artificial intelligence began to reshape, Leskovec Luck It was shaken by the explosion of the field to the main current. Stanford said that he had a prestigious computer science program, which he felt that he felt that he saw before his future was born, even before the future was born, but the release of GPT-3 to the public was shaken.
Leskovec said, “A few years ago, a big one is not clear what our role was in this world, I don’t know, we had existential crisis,” he said.
“The breakthroughs in AI will do research only for us, so what will we do?” He said that he spent a lot of time by talking to the doctoral students about how to organize themselves even about what their roles will be in the world. “Existential” and “amazing” he said. Then, another surprise received: the request for provoking by the student for the change in the test.
“He got out of the group,” he said, especially teaching assistants, undergraduate students of the previous generation of computer science. Their idea was simple: “We do a paper exam.”
As a catalyst for change
Leskovec, a leading researcher in Graphic Data in Stanford and AI applications in Biology, explained Pivoti with a surprise and thoughtful mixture. Historically, the courses relied on open books and home exams where students could benefit from the textbooks and the internet. They couldn’t use others’ code and solutions, but the rest was a fair game. Openai’nin GPT-3 and GPT-4 large language models exploded at the scene, students and teaching assistants began to question whether the evaluations should be handled differently.
Now he said that he had a lot more work for him and Tas, and he said that these exams took “much longer” to take notes. However, they all acknowledged that this was the best way to test student knowledge. For Leskovec, an AI veteran, AI’s age surprised him by putting a higher workload for himself and other people. He also said that AI has created a new “additional work”, as well as being “less trees in the world ,, from all the papers he printed. He felt like a audience in a “rock concert ,, but insisted that he did not return to AI to synthesize and analyze all exams.
“No, no, no, a hand note,” he insisted.
A student -guided solution
Leskovec’s solution, violent debate Common fraud reports about how AI has changed higher education led to many colleges to prohibit the use of AIs. Other professors return to the article exam and Famous Blue Books One of the 90s high school memories. A Professor of a New York University He even suggested to adopt the old test forms such as oral and written examination. In the example of Leskovec, the solution of AI professor for AI age is to get away from AI for the test.
When asked if he was worried about the students who cheated with AI, Leskovec asked another question: iz If you allow a calculator in your math exam and say that the calculator is not allowed, you will have a different test. ” AI, which resembles an artificial intelligence calculator, said that it was an incredibly powerful tool that çıkar It has revealed all of us and surprised us all ,, but at the same time “Very flawed… We need to learn how to use this tool and we should be able to test the vehicle and think of people on their own.”
What is AI skill and what is human skill?
Leskovec wrestle with a question that touches everyone of the labor force: What is a human skill, what is AI skill and where do they unite? MIT Professor David Autor and Google SVP James Manyika Discussed Atlantic Vehicles such as a calculator or artificial intelligence usually enter two buckets: automation and cooperation. Think of the dishwasher or word processor on the one hand. The cooperation tool “requires human participation” and the problem with artificial intelligence “is that it does not go properly to both of them [bucket]. “
The labor market sends a response to the AI application, which is equal to something like a response. Magic 8 Ball: “Female reply. Try again later.” The Federal Affairs Report has recently grew anemic since spring. Disappointing Expectations With only 22,000 business pressure in August. Most economists attribute the lack of recruitment to the uncertainty about President Donald Trump’s tariff regime. More than one court decided illegal And it looks like going appearance Supreme Court. However, the AI application does not go smoothly at the institutional level with a MIT work (not depending on Autor). 95% of productive AI pilots failSoon he watched with a Stanford work. The beginning of a collapse in recruitment At the entry level, especially in automation by AI.
For another perspective Free Sunday Upwork The opening launched a monthly recruitment report and revealed which full -time works are rewarded by the market. The answer is that “AI skills ver is highly demanded, and even if companies do not hire full -time employees, they become high -wage and highly qualified free labor.
In spite of a softer labor market, UPWORK grows 31% of companies’ strategically flexible abilities to address the temporary gaps of the labor force ”and large businesses grow 31% of UPWORK’s high -valuable jobs on the platform (more than $ 1,000). Smaller and medium -sized enterprises, artificial intelligence and machine learning demand by 40%. However, Upwork also sees the demand for increasing types of skills: a person who is good to cooperate with AI.
UPWORK says that the AI is the most creative and has raised expertise in higher valuable studies, which are visible among the categories of design, writing and translation. One of the best skills hired in August was to control the truth, considering the need for the verification of the AI outputs to the verification of human beings ”.
Kelly Monahan, General Manager of the Upwork Research Institute, said, “People are returning immediately”.
“In fact, we see that human skills come primarily,” he said, he thinks that people noticed AI hallucinates over too much time to replace human participation. “I think people see what they see, because they now use the content created by AI, they need real control.”
Monahan, who expanded this line of thought, said that the developing landscape of “AI skills” is more and more valuable. Legal is a growing category in August, he said, he said that legal expertise should be emphasized to control the legal article created by AI. If you do not have advanced skills in a specific field, the content created by AI is “easy to take” and businesses hire to protect against it.
Leskovec agreed to face the skill gap that seems to be facing the entry -level workers trying to hire on the one hand and that he was facing companies struggling to implement AI effectively.
“I think it almost has to do it again. Human expertise is much more important than ever. [before]. “The essence of the problem is the essence of the problem,” because how do young workers need to get the field expertise needed to effectively cooperate with AI?
Leskovec has a role that colleges will play, but the organizations said it was. He asked a rhetoric question: If they take the time to take the time to take and train them, how should they have senior talented workers?
When asked Luck In order to examine the landscape and to evaluate that we are currently using artificial intelligence, Leskovec, as professors and workers, said that we were “too early in this regard. He said he thought we were in the “Future Stage with Solution”. A professor who finds news ways to check the knowledge of his students and solutions such as a hand -grade exam and professor.
This story took place at the beginning Fortune.com
