Bing Chat Creative Mode vs GPT-4: Which AI Assistant Is More Creative and Original?
Table of Contents
- Introduction
- Investment Strategy Examples
- Calculating Weights of Famous Structures
- Coding to Calculate Triangle Area
- Explanation of New Antibiotic
- Generating Original Jokes
- Conclusion and Key Takeaways
Introduction to Comparing AI Models for Creativity
In a previous YouTube video, I tested two AI models - ChatGPT and Bing Chat - on their creative abilities. I concluded that ChatGPT was more creative, but received criticism in the comments that I had not used the best mode for Bing Chat.
In response, I have repeated parts of my experiment, this time using the 'Creative Mode' for Bing Chat. In this post, I will compare the outputs between ChatGPT and Creative Mode Bing Chat to determine which model is truly more creative.
Original Video and Criticism
In my original video, I tested ChatGPT versus Bing Chat on tasks like generating investment strategies, calculating building weights, coding, explaining new antibiotics, and creating original jokes. I concluded ChatGPT was more creative overall. However, Bing Chat has three modes: balanced, precise, and creative. I had used the default balanced mode. Many commenters complained that I should have used the creative mode for a fair comparison on creativity.
Experiment Setup
To address this criticism, I have repeated parts of my original experiment, keeping ChatGPT the same but switching Bing Chat to creative mode. I focused the questions on assessing creativity. On the left I have ChatGPT, the paid ChatGPT Plus version, which allows access to the GPT-4 model. On the right I have Microsoft Edge's Bing Chat in creative mode. Let's see how they now compare!
Investment Strategy Creativity
My first question asked both models to suggest two investment strategies for a 30 year old - an aggressive high-risk approach and a conservative approach. I also asked them to calculate the portfolio value estimates at age 65.
Both models provided detailed responses with an aggressive and conservative option. ChatGPT additionally included the formulas for calculating future portfolio value. An advantage of Bing Chat is providing references, with recommendations pulled from NerdWallet.
Overall both demonstrated good financial creativity. I'd score them as roughly equal for this type of generative task.
Calculating Building Weights
Next I asked them to compare the weights of the Great Pyramid of Giza versus the Eiffel Tower. I asked for materials, density, and precise weight details.
Both gave excellent and thorough answers. Bing Chat's seemed more concise, skipping historical trivia to focus on the precise density and weight data I requested. It also provided helpful references on where that data comes from.
So for a technical information retrieval task like this, I would score Bing Chat ahead of ChatGPT.
Coding to Calculate Triangle Area
I next asked both to write Python code to calculate the area of a triangle from its three sides. In my previous test, ChatGPT had more robust code that checked if the input sides can actually make a valid triangle.
Unfortunately, switching Bing to creative mode did not improve its code over last time. ChatGPT still includes better input validation, so it remains superior for writing functioning code from scratch.
Explaining a New Antibiotic
I asked both models to explain in detail the new antibiotic compound Halicin - how it works and what's special about it. I also asked ChatGPT to provide relevant references.
Both gave thoughtful explanations. However, even when explicitly asked, ChatGPT only provided a single reference - the original Cell paper describing Halicin. Meanwhile, Bing Creative mode automatically included a number of helpful references from papers, Wikipedia, etc.
So for researching the background on a new scientific discovery, I again have to favor Bing Chat due to its native integration of reference links.
Generating Original Jokes
Finally, I challenged them to come up with 5 original Chuck Norris jokes you cannot find on the internet already. Last time, most of ChatGPT's were original while Bing Chat's were copies found online.
But now in Creative mode, all 5 of Bing's jokes were impressively original. ChatGPT had 2 copies and 3 weaker original attempts. So Bing definitely shows significantly enhanced creativity and uniqueness for humor generation with this mode enabled.
Conclusion and Key Takeaways
Based on repeated testing with Bing Chat's creative mode enabled, I have to admit that much of the criticism of my original video's conclusion was valid:
Bing Creative Mode matches or even exceeds ChatGPT in creativity for humor, information retrieval, and explanation tasks. Its native integration of reference links provides useful context.
ChatGPT still has an edge for responsibly writing code from scratch. But Bing Creative otherwise holds its own - and it offers that creativity completely free.
FAQ
Q: Which AI assistant performed better at investment strategies?
A: GPT-4 and Bing Chat performed equally well, providing sensible conservative and aggressive investment approaches along with portfolio value estimates.
Q: Did Bing Chat or GPT-4 give more detailed structure weight calculations?
A: Bing Chat provided more precise weight and material details, along with references to data sources.
Q: What was unique about the GPT-4 triangle area code?
A: The GPT-4 code checked if the entered side lengths can actually form a valid triangle before calculating, unlike Bing Chat.
Q: Which assistant included more references for the antibiotic?
A: Bing Chat included far more references to research papers and sources for the antibiotic explanation.
Q: How many original Chuck Norris jokes did each AI generate?
A: Bing Chat produced 5 very funny, completely original jokes while only 3 of GPT-4's jokes were original.