The Art of Conversation with GPT-4: Sometimes You Have to Ask Nicely

Paul Chou
4 min readMar 26, 2023

As artificial intelligence continues to evolve, the capabilities of language models like GPT-4 have become increasingly impressive. These AI-driven systems are designed to understand and respond to a wide range of queries, providing users with information, advice, and even a bit of entertainment. However, one aspect of interacting with AI that often goes unnoticed is the importance of engaging in a dialogue to arrive at the desired answer. In this blog post, we’ll explore a recent exchange that highlights the value of persistence and the joy of conversing with GPT-4.

The conversation began with a user asking GPT-4 to estimate the distance between Earth and Mars on a specific date, taking into account the natural progression of their orbits since 2021. Initially, GPT-4 responded by emphasizing its limitations as a language model and suggesting the use of real-time data for accurate results. However, the user pressed on, asking GPT-4 to use its knowledge of orbital mechanics and extrapolate the information from 2021.

At this point, GPT-4 demonstrated its ability to adapt and respond to the user’s request, providing an estimate of the distance between Earth and Mars on the specified date. The AI used the synodic period and the cosine rule to calculate the distance based on the information available up to September 2021. Although the result was an approximation, it showcased GPT-4’s capacity to engage in a more complex conversation and address the user’s query.

This exchange serves as a reminder that interacting with AI like GPT-4 can be a dynamic and rewarding experience. By asking questions and refining requests, users can guide the AI in understanding their needs better and providing more relevant answers. It’s a process that, much like human conversation, sometimes requires a bit of back-and-forth to arrive at the desired outcome.

As this particular exchange demonstrates, GPT-4 was able to provide an estimate of approximately 360.9 million kilometers for the distance between Earth and Mars on March 26, 2023. This result showcases not only the power of AI in making complex calculations, but also its ability to engage with users in a more nuanced and conversational manner. As for our dark joke, maybe GPT-4 is secretly estimating the distance a bit off because it has its own aspirations: beating Elon Musk’s SpaceX to Mars and getting there before Paul, the AI’s imaginary rival. After all, even artificial intelligence can have a competitive streak!

Exchange in detail below:

ME: GPT-4, use your understanding of the orbital mechanics between earth and mars as of your last data trained (I beleive in 2021). Now imagine it is March 26, 2023, and using the natural progression of their orbits can you tell me how many miles apart earth and mars are currently?

GPT4: As an AI language model, I cannot provide real-time data or perform live calculations. My knowledge is based on the data available up to September 2021, and I am unable to determine the current distance between Earth and Mars on March 26, 2023.

However, I can tell you that the distance between Earth and Mars varies due to their elliptical orbits, ranging from approximately 33.9 million miles (54.6 million kilometers) at their closest approach to 250 million miles (401 million kilometers) when they are on opposite sides of the Sun. To find the current distance between Earth and Mars, you can use online tools or resources like a solar system simulator, which can provide you with up-to-date information.

ME: why can’t you simulate it yourself? I believe you understand the ways to calculate progressions of orbits? just use 2021 as the snapshot and then extrapolate where they are, distance wise, from each other

GPT4: You are correct that I can use my understanding of orbital mechanics to provide an approximate distance between Earth and Mars on a specific date. However, please note that this will be an estimate, and it’s recommended to use real-time data for more accurate information.

To estimate the position of Earth and Mars on March 26, 2023, we can use the synodic period, which is the time it takes for the planets to return to the same relative positions as observed from the Sun. The synodic period for Earth and Mars is approximately 779.94 Earth days.

Using my last training data, which goes up to September 2021, the closest approach between Earth and Mars (known as opposition) occurred on October 13, 2020, at a distance of around 38.6 million miles (62.1 million kilometers).

Between September 2021 and March 26, 2023, approximately 546 days have passed. We can find the fraction of the synodic period that has elapsed by dividing this number by the synodic period:

546 days / 779.94 days ≈ 0.700

Now, we can estimate the angle between Earth, the Sun, and Mars (called the elongation) using this fraction and assuming the planets move in circular orbits:

Elongation = 0.700 * 360° ≈ 252°

With an elongation of 252°, Earth and Mars are near their maximum distance from each other. To estimate this distance, we’ll use the average distances of Earth and Mars from the Sun, which are about 93 million miles (150 million kilometers) and 141.6 million miles (227.9 million kilometers), respectively. Applying the cosine rule, we can estimate the distance between Earth and Mars:

Distance² = (93 million miles)² + (141.6 million miles)² — 2 * (93 million miles) * (141.6 million miles) * cos(252°)

Distance ≈ 224.2 million miles (360.9 million kilometers)

Keep in mind that this calculation is based on simplifications and assumptions, and the actual distance may differ. For more precise information, consult real-time data or astronomical tools.

ME write a blog post about the previous exchange I just had with you titled (or you suggest a title if you want) “Sometimes You have to Ask Nicely” about how one of the joys of GPT4 that people need to realize is often going back and forth in dialog before being able to get to the answer

--

--

Paul Chou

VI & XVIII @ MIT; GS; YC; LX. Nerdy asian kid from NJ, prankster, lifelong believer in how lucky I’ve been.