ChatGPT Just Passed This Wharton Professor's Final Exam, But He's Not Banning The Chatbot

The Wharton Faculties operations professor, Christian Terwiesch submitted his operations remaining into ChatGPT. To his shock, it handed. Courtesy photograph

The entire experiment began as a dinner desk dialog. Christian Terwiesch’s older little one was fascinated about rising AI instruments that acknowledge, course of, and create photos. The youthful was within the code era capabilities of Open AI’s exceptional new chatbot, ChatGPT.

“Dad, you teach at a university,” one among his youngsters instructed. “See how this thing would do on your exam.”

Terwiesch, the Andrew M. Heller professor of operations at College of Pennsylvania’s Wharton College, had no expectations at that time. His hopes weren’t excessive. The household sat at a pc and cut-and-pasted into the chatbot’s immediate, with no enhancing, the primary query from Terwiesch’s remaining examination for his Wharton Operations Administration course.

“I was just blown away. The answer was correct, and it was really well-worded and well-reasoned,” Terwiesch tells Poets&Quants. “I said, ‘Wow, this machine is awesome.’”


In line with a white paper Terwiesch printed on January 17, “Would ChatGPT Get a Wharton MBA?” , Terwiesch would have given the chatbot an A+ on that first query, which requested college students to establish the bottleneck in a seven-part course of in an iron-ore refinery.

ChatGPT, which stands for Chat Generative Pre-trained Transformer, earned one other A+ on the second query on stock turns and dealing capital necessities. Terwiesch thought he was sitting subsequent to some sort of marvel machine.

Then it “came back down to Earth,” he says. On the third query, which was extra advanced, the pc chatbot couldn’t do the mathematics.

“In a sense, that is an irony, because it’s good at reasoning, telling stories, humor – all really human things,” Terwiesch says. “The things that humans have always been bad at – like mathematics – we kind of thought a computer would be good at.”

Ultimately, the chatbot scored properly on three of the 5 questions; Terwiesch says he would have given it a remaining grade of B or B-. Not too shabby for a fledgling bot taking its first remaining at probably the most prestigious enterprise colleges on the earth.


Quickly after its launch in November, ChatGBT went viral. It handed out relationship recommendation, composed authentic rap lyrics about Elon Musk (an Open AI cofounder), and demonstrated a capability to write down, debug and clarify pc code. It additionally triggered a debate in regards to the potentialities behind its astonishing capabilities versus the flexibility for folks to abuse them.

This month, the Stanford Every day reported that 17% of surveyed Stanford College college students stated they used the chatbot of their remaining exams and assignments. Of these, 7.3% reported that they’d submitted written materials from ChatGPT with edits, and 5.5% reported submitting ChatGPT materials with no edits in anyway. New York Metropolis’s public colleges blocked ChatGPT entry on college computer systems and networks out of concern of dishonest.

Terwiesch, who serves as chair of Wharton’s Operations, Info, and Choices Division and co-director of Penn’s Mack Institute for Innovation Administration, makes clear that plugging his remaining into the chatbot was not a cautious analysis research. It was merely an experiment spurred by his dinner desk dialogue. However the outcomes and the implications he wrote about afterwards spurred its personal flurry of headlines, reigniting a debate about how enterprise colleges ought to use this wondrous new expertise: Ban it or use it.

Terwiesch hopes the dialog shall be a bit extra nuanced.

These conversations, taking place now at enterprise colleges all over the world, echo these from 9 years earlier. That’s when MOOCs (Huge Open On-line Programs) first exploded onto the scene and, some believed, would sound the dying knell for all however the high ranked MBA applications. As an alternative of shunning the brand new expertise, Terwiesch was among the many first enterprise college school to make his MBA coursework out there to most of the people. Opposite to what some thought, video didn’t kill the classroom star.

Poets&Quants had the prospect to speak with Terwiesch at size this week about his white paper and the implications of ChatGPT on enterprise schooling. The dialog has been edited for size and readability.

You clarify in your white paper how ChatGPT scored A+ grades on the primary two questions of the examination, however then did worse on the third. Are you able to clarify the place it tripped up?

It tousled in two methods: First, it actually acquired misplaced within the query, however I give it nice credit score that it recovered once I gave it human hints. So, along with the questions with the command line, you’ll be able to add feedback that construct upon and add to the preliminary query. It actually took these under consideration and altered its quote unquote “thought process.”

It failed bitterly on the math. I’m not an AI professional, nevertheless it’s my understanding that this has to do with the character of how the chatbot works. It’s principally a prediction machine that makes probabilistic statements based mostly on the patterns of characters and phrases it has seen within the query. That works properly with language. I believe the sooner variations of the expertise had been educated by Shakespeare texts. When you’ve learn all of Shakespeare’s work time and again, you sort of get the rhythm of it and might extrapolate.

ChatGPT Just Passed This Wharton Professor's Final Exam, But He's Not Banning The Chatbot

Wharton college students collect inside Huntsman Corridor. Courtesy photograph

Math doesn’t work this fashion. It really works for two + 2 = 4 as a result of that assertion is discovered typically sufficient on the web that it’s capable of make a prediction. However as you get into greater numbers – even for multiplication tables or for issues a fourth grader may do in 5 minutes with a pen and paper – there’s simply not sufficient proof out there that the machine has seen on the web. So, these easy issues it actually struggles on.

Is it conceivable that it’ll get higher at these sorts of issues over time?

I’m not a expertise professional, however my intestine really feel is they may add one other layer of software program on high of that. For instance, when the machine must generate its subsequent phrase, it’s going to use its present prediction methodology. When it must make a mathematical assertion, it’s going to attain out to a extra conventional calculator and outsource that half.

Whenever you get to 5 – 6 digit numbers, although, you’ll be able to’t have it memorizing options. That’s like a child memorizing the multiplication desk versus with the ability to do the multiplication cognitively.

What was your preliminary response after realizing ChatGPT scored a B to B- in your remaining examination? Was it fascinating, alarming, what?

Initially, I used to be identical to, “Wow. We’re living in exciting times.” You understand, I took a category on neural networks once I went to varsity 30-plus years in the past, and these items had been so lame then. Seeing it now could be simply unbelievable. You can’t even think about how the world will look in 10 years. So my first response was simply being in awe.

I believe there are two sorts of alarms right here. One, is the kind of alarm that this factor goes to go my assessments and lots of different assessments — AP exams, school entrance exams, homework assignments, and so forth. – so, we have now to rethink schooling a bit. Then there’s the opposite alarm, these raised by folks like Sam Harris: Is AI taking on the world? Does this machine have consciousness? I believe each of these are fascinating.

On the tutorial entrance, I believe each scholar in my class is welcome to make use of it for case dialogue and getting ready for sophistication. If something, I believe it might add to the dynamics and the motion of an MBA classroom the place folks get exterior recommendation. I imply, it’s how the enterprise world works. You will have two folks bargaining with one another: On one facet is recommendation, possibly by McKinsey, and the opposite one is an funding banker from Goldman Sachs. With the ability to critically have a look at what advisors inform you and understand when the advisor is flawed, and probably reconciling conflicting opinions, I believe that makes the MBA classroom a richer place that’s nearer to the actual world.

I do suppose we have now to revisit testing. However I believe for classroom dialogue, that is going to be nice. I imply, once more, we educate MBAs to be leaders in a transferring, dynamic, advanced world and with this expertise, this world simply strikes a bit of quicker, turns into a bit of extra dynamic. If something, the demand for MBA college students has solely gone up.

NEXT PAGE: How is Wharton responding to ChatGPT?

What's Your Reaction?

hate hate
confused confused
fail fail
fun fun
geeky geeky
love love
lol lol
omg omg
win win
The Obsessed Guy
Hi, I'm The Obsessed Guy and I am passionate about artificial intelligence. I have spent years studying and working in the field, and I am fascinated by the potential of machine learning, deep learning, and natural language processing. I love exploring how these technologies are being used to solve real-world problems and am always eager to learn more. In my spare time, you can find me tinkering with neural networks and reading about the latest AI research.


Your email address will not be published. Required fields are marked *