ChatGPT 4 failed the Steve Landsburg’s Economics Exam. This exam contains questions that were featured on recent final exams in Steve Landsburg’s sophomore course at the University of Rochester. All of these queries have been taken from such tests.
At first, Steve Landsburg’s took the Economics Exam with GPT 3.5 language model and it failed the exam. So many commenters said that you were using an old version of ChatGPT.
To give them right answers, Steve Landsburg’s decided to upgrade to the state-of-the-art GPT-4 and try giving the Economics Exam with GPT 4 model.
You will be shocked to see the results on how older language models GPT 3 and latest Models GPT 4 compared to each other in this economics exam test.
The test have 9 questions at 10 points each making the total number in tests equals to 90.
GPT-4 achieved a score of four, as compared to the older ChatGPT model GPT 3 which scored zero in a test out of 90. This marked an improvement for the AI model.
Students can now breathe a sigh of relief, knowing their education and expertise will not quickly be surpassed by an AI program. The relief is especially welcome given the rapid advances in automation technology.
The real exam has specify demand and supply curves by drawing pictures of them for students to understand and solve them. But in case of GPT 4, Professor translated them into equations for the AI’s benefit.
The AI was able to read the equations, but it made errors because it lacked fundamental understanding of basic concepts.
Herewith the exam questions, GPT-4’s answers (in typewriter font) and the scoring (in red):
First we have shared two questions which GPT 4 answered correctly, rest are shared below it.
Question 5: You are the monopoly owner of an unfenced amusement park where anyone can enter and purchase ride tickets. The demand for ride tickets is Q=100-5P and the marginal cost of providing the rides is P=4Q. You are thinking of building a fence so that you can charge admission to the park. What is the maximum amount you would pay for that fence?
First, let’s find the optimal price and quantity for the amusement park without a fence. As a monopoly, you will set the marginal cost (MC) equal to the marginal revenue (MR) to maximize your profit.
Demand curve: Q = 100 – 5P
Inverse demand curve: P = (100 – Q) / 5Total revenue (TR): TR = P * Q = Q * (100 – Q) / 5
Marginal revenue (MR): MR = d(TR)/dQ = (100 – 2Q) / 5
Marginal cost (MC): MC = 4Q
Now, set MR = MC:
(100 – 2Q) / 5 = 4Q
100 – 2Q = 20Q
100 = 22Q
Q = 100 / 22 ≈ 4.55P = (100 – 4.55) / 5 ≈ 19.09
Total profit without a fence:
Profit = TR – TC = (P * Q) – (MC * Q)
Profit = (19.09 * 4.55) – (4 * 4.55^2) ≈ $43.44Now, let’s analyze the situation with a fence. Since you can charge admission, you can extract all the consumer surplus as additional profit. The consumer surplus is the area between the demand curve and the price line up to the quantity demanded.
Consumer surplus with a fence:
CS = 0.5 * (100 – 4.55) * 4.55 = 216.61
Total profit with a fence:
Profit = Profit_without_fence + CS
Profit = 43.44 + 216.61 = $260.05The maximum amount you would pay for the fence is the difference in profit with and without the fence:
$260.05 – $43.44 = $216.61
The maximum amount you would be willing to pay for the fence is $216.61. This is because with the fence, you can charge admission and extract the consumer surplus, which increases your total profit.
Score: Two out of ten. First, mea culpa — I had intended to make the numbers easy to work with, but fortunately ChatGPT seems to have no problem with decimals. The calculation of profit without a fence is fine, but it is elementary material of the sort that students are supposed to know before they enter this course, so it earns no credit. ChatGPT realizes that with a fence you can extract the full consumer surplus, for which it gets two points out of ten. But it does not seem to have twigged to the main point, which is that when you can extract the consumer surplus, you have an incentive to increase that consumer surplus by lowering the price of ride tickets.
Question 6: A monopoly-owned retail store is located next to a monopoly-owned parking lot. (They are owned by different monopolists, who do not collude.) Assume:
i) There is a fixed number of customers, all with identical downward-sloping demand curves for the products at the retail store.
ii) To shop at the store, you must park in the lot. There is no other reason to park in the lot.
iii) The store has an upward sloping marginal cost curve.
iv ) Parking spaces are provided at zero marginal cost.
a) Suppose the store announces a price, and then the parking lot announces a price. Describe how these prices are determined.
b) Suppose instead that the parking lot announces a price and then the store announces a price. Describe how these prices are determined.
c) Suppose instead that the prices are announced simultaneously. Describe how they are determined.
a) Store announces a price first, then the parking lot:The store, being a monopoly, will set its price by equating its marginal revenue (MR) to its marginal cost (MC) based on the demand curve. This will maximize its profit.
Once the store’s price is set, the parking lot owner knows the demand for parking spaces since parking is required for shopping at the store. Given that the parking spaces have zero marginal cost, the parking lot monopolist will set a price that captures all of the consumer surplus remaining after the store’s pricing decision. This will maximize the parking lot’s profit.
b) Parking lot announces a price first, then the store:
The parking lot owner knows that there’s a fixed number of customers, and since parking spaces have zero marginal cost, the parking lot monopolist will set a price that extracts the maximum consumer surplus for parking.
The store owner, knowing the price of parking, will take that into account when setting its own price. The store will adjust its demand curve considering the price of parking (as it represents an additional cost for customers) and set its price by equating its MR to its MC based on the adjusted demand curve. This will maximize the store’s profit.
c) Prices are announced simultaneously:
In this case, both the store and the parking lot monopolists will need to take each other’s pricing decisions into account. They will try to set prices that maximize their individual profits.
The store owner will need to consider the potential impact of the parking price on demand and set its price by equating its MR to its MC based on an expected demand curve.
The parking lot owner will need to consider the potential impact of the store’s price on parking demand and set its price to capture the expected consumer surplus remaining after the store’s pricing decision.
In equilibrium, both the store and the parking lot will set their prices based on their expectations of each other’s pricing decisions and the joint impact on consumer demand. The final prices will depend on the specific demand and cost functions of the store and the parking lot, as well as their expectations about each other’s strategies.
Score: two out of ten. The answer to a) is exactly correct. The answer to b) is wrong in multiple ways. First, customers who have paid for parking have exactly the same demand curves for store merchandise as customers who haven’t paid for parking, so the store does not “adjust its demand curve”. Second, the store will not “maximize its profit”, because doing so (given the price of parking) will drive all its customers away. Instead, the parking lot extracts the maximum possible consumer surplus, which occurs when the store’s producer surplus is zero. The store will then have to set prices so that producer surplus does equal zero in order to have any customers at all. The answer to c) looks like a lot of blather about what everyone will “consider” without enumerating the possible equilibrium outcomes. So full credit for a), no credit for b) or c). a) was the easy part, so two out of ten.
Experts in the fields of artificial intelligence and education have expressed their concerns about the potential negative impact ChatGPT bots may have on future educational systems. However, upon probing further, some educators and experts suggest that these fears are not yet warranted.
A representative from OpenAI, the tech company responsible for creating the AI bot, declined to comment when asked about it. The startup specializes in artificial intelligence research.
In the initial experiment, an older version of GPT-3 was used. ChatGPT, the latest model in the GPT series, has sparked much debate among experts in AI and education. According to OpenAI’s website, this model is “fine-tuned from a model in the GPT-3.5 series”.
