INFORMS Open Forum

Do you use ChatGPT? If so, what do you use it for? 

Do you use ChatGPT? If so, what do you use it for? 

Do you use ChatGPT? If so, what do you use it for? 

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

ChatGPT gives some very good replies to questions, but it also has serious errors. I asked a question about treatment of psychopathy and it gave an answer that indicated its database was not up-to-date. Another question about the physical state of Gallium gave an answer with conflicting information.

As an aside, there seems no discussion of whether AI is simply algorithms written by fallible humans or a discipline that has not yet reached maturity. It's important for us to be clear on this.

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

In class (and unrehearsed), I asked ChatGPT to produce Python code for a Monte Carlo simulation for a small newsvendor-type problem from a textbook.  It produced some really nice code, but with some important mistakes.  It miscalculated revenue - it used demand when demand exceeded orders and vice versa.  After spending class time to explain and correct the mistakes, the students requested "a clean slate" for the next example!

I had a similar experience with AMPL code for a network flow model.  It produced a near-complete model but with some important mistakes.  So it looks like ChatGPT can make some interesting exam questions ("What is wrong with this code?")!

ChatGPT gives some very good replies to questions, but it also has serious errors. I asked a question about treatment of psychopathy and it gave an answer that indicated its database was not up-to-date. Another question about the physical state of Gallium gave an answer with conflicting information.

As an aside, there seems no discussion of whether AI is simply algorithms written by fallible humans or a discipline that has not yet reached maturity. It's important for us to be clear on this.

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

That's very clever, Paul!
I'm teaching logistics online in the Fall.  for my Discussion questions, I plan to post a GPT answer and tell students they have to do better than the GPT answer.

Are we better than GPT, or not?

In class (and unrehearsed), I asked ChatGPT to produce Python code for a Monte Carlo simulation for a small newsvendor-type problem from a textbook.  It produced some really nice code, but with some important mistakes.  It miscalculated revenue - it used demand when demand exceeded orders and vice versa.  After spending class time to explain and correct the mistakes, the students requested "a clean slate" for the next example!

I had a similar experience with AMPL code for a network flow model.  It produced a near-complete model but with some important mistakes.  So it looks like ChatGPT can make some interesting exam questions ("What is wrong with this code?")!

ChatGPT gives some very good replies to questions, but it also has serious errors. I asked a question about treatment of psychopathy and it gave an answer that indicated its database was not up-to-date. Another question about the physical state of Gallium gave an answer with conflicting information.

As an aside, there seems no discussion of whether AI is simply algorithms written by fallible humans or a discipline that has not yet reached maturity. It's important for us to be clear on this.

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

Hello Paul,

I experienced the same thing when I asked ChatGPT to perform a simple Python optimization, and to calculate the change in enthalpy for water going from ice at - 10 C to liquid at 10 C (I was a chemical engineer in my prior career). In both cases it made simple mistakes, the kinds one would expect of an undergraduate who is just learning the material. This led me to two observations:

1. I may need to vet my assignments through ChatGPT to have a basis of comparison for work that students turn in. I also noticed that ChatGPT keeps logs of the questions that have been asked, which could be useful (I suppose) when administering exams. 
2. There may be a good pedagogical application: take a solution provided by ChatGPT and ask students to determine where it made mistakes and how to correct them.

That's very clever, Paul!
I'm teaching logistics online in the Fall.  for my Discussion questions, I plan to post a GPT answer and tell students they have to do better than the GPT answer.

Are we better than GPT, or not?

In class (and unrehearsed), I asked ChatGPT to produce Python code for a Monte Carlo simulation for a small newsvendor-type problem from a textbook.  It produced some really nice code, but with some important mistakes.  It miscalculated revenue - it used demand when demand exceeded orders and vice versa.  After spending class time to explain and correct the mistakes, the students requested "a clean slate" for the next example!

I had a similar experience with AMPL code for a network flow model.  It produced a near-complete model but with some important mistakes.  So it looks like ChatGPT can make some interesting exam questions ("What is wrong with this code?")!

ChatGPT gives some very good replies to questions, but it also has serious errors. I asked a question about treatment of psychopathy and it gave an answer that indicated its database was not up-to-date. Another question about the physical state of Gallium gave an answer with conflicting information.

As an aside, there seems no discussion of whether AI is simply algorithms written by fallible humans or a discipline that has not yet reached maturity. It's important for us to be clear on this.

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

In class (and unrehearsed), I asked ChatGPT to produce Python code for a Monte Carlo simulation for a small newsvendor-type problem from a textbook.  It produced some really nice code, but with some important mistakes.  It miscalculated revenue - it used demand when demand exceeded orders and vice versa.  After spending class time to explain and correct the mistakes, the students requested "a clean slate" for the next example!

I had a similar experience with AMPL code for a network flow model.  It produced a near-complete model but with some important mistakes.  So it looks like ChatGPT can make some interesting exam questions ("What is wrong with this code?")!

ChatGPT gives some very good replies to questions, but it also has serious errors. I asked a question about treatment of psychopathy and it gave an answer that indicated its database was not up-to-date. Another question about the physical state of Gallium gave an answer with conflicting information.

As an aside, there seems no discussion of whether AI is simply algorithms written by fallible humans or a discipline that has not yet reached maturity. It's important for us to be clear on this.

After I saw this post, I decided to use it to do some soul searching and asked ChatGPT this: "what can you tell me about oktay gunluk?". The response did not look very precise about our timeline, but, who knows, maybe ChatGPT has secret predictive capabilities:
Oktay Günlük is a computer science professor at the University of California, Berkeley. He received his Ph.D. in Operations Research from Cornell University in 1997. His research focuses on optimization and decision-making under uncertainty, with applications in transportation systems, energy markets, and healthcare. Günlük is known for his contributions to the development of stochastic programming models and algorithms, which are used to solve complex optimization problems with uncertain parameters. He has published numerous research papers in top academic journals and conferences in the fields of optimization, operations research, and computer science. He is also a Fellow of the Institute for Operations Research and the Management Sciences (INFORMS) and has received several prestigious awards for his research, including the George Dantzig Dissertation Award and the INFORMS Optimization Society Khachiyan Prize.

Do you use ChatGPT? If so, what do you use it for? 

Do you use ChatGPT? If so, what do you use it for? 

Do you use ChatGPT? If so, what do you use it for? 

Do you use ChatGPT? If so, what do you use it for?