I tried using gpai | Testing how well AI can solve STEM problems
How I found out about gpai
Recently, it has become common to use generative AI like ChatGPT and Claude in our daily work.
It was in this context that I discovered gpai.
The fact that it is an "AI specialized in STEM (Science, Technology, Engineering, and Mathematics)" caught my interest, so I decided to test if it could handle problems in the field of optics that I work with regularly.
Mathematics and physics often involve many diagrams and formulas, which can sometimes be challenging for general-purpose chat AIs.
Therefore, I decided to test just how practical this specialized AI really is.
What is gpai Problems?
gpai Problems is an AI feature that analyzes STEM problems, including those with formulas and diagrams, and explains the solution step-by-step.
You can not only input problem text but also upload images for analysis, and it will show you the intermediate steps and the reasoning process. Additionally, it offers features to ask follow-up questions after the answer is provided or to generate similar problems.
I actually tried having it solve an optics problem
This time, I input a basic problem from Fourier optics.
Once I entered the problem text, the answer was generated in a few dozen seconds.
What surprised me first was that instead of just giving the answer immediately, it explained it in the following flow:
Given conditions
Formulas used
Calculation steps
Explanatory diagrams
Final analysis
.
Because it even supplements the explanation of "why that formula is used," I felt it was easy to use not just for checking answers, but also for learning purposes.
Since you can understand by combining diagrams and text, it seems useful for presentation materials and studying as well.
Good points I felt after using it
1. Relatively smooth even with problems containing mathematical formulas
While general chat AIs can struggle with inputting mathematical formulas or handling diagrams, gpai is designed with science and engineering in mind, so the flow of handling formulas and diagrams was natural.
2. It explains "why it turns out that way"
It shows not just the answer, but also the way of thinking and intermediate steps, so it is useful in situations where you want to deepen your understanding.
I felt it was also suitable for exam preparation and review.
3. Easy to ask follow-up questions
"What is the meaning of this formula?"
"Is there another way to solve this?"
"What happens if the conditions are changed?"
Because you can continue asking questions like these, it was convenient to be able to deepen your understanding in a single interaction. However, be aware that there are limits to credits in the free plan.
Points of concern
Of course, it is not omnipotent.
For specialized content at the graduate school level or problems that depend on experimental conditions, it is important to verify the answers yourself rather than adopting them as they are.
Also, I had the impression that the more specific the problem statement you input, the more accurate the answer you get.
Recommended for people like this
I felt that gpai is suitable for the following types of people.
Science and engineering students
Those studying mathematics, physics, or engineering
Those who want to streamline learning and document creation in the laboratory
Those who want to consult AI about problems involving mathematical formulas and diagrams
In particular, I think it is an easy-to-use service for those who want to "understand the way of thinking" rather than just "knowing the answer."
Summary
What I felt after actually using gpai was its ease of use unique to an AI specialized in STEM fields.
Its ability to handle problems involving mathematical formulas and diagrams, step-by-step explanations, and the ease of asking follow-up questions are strengths that differ slightly from general chat AIs.
Of course, final verification of the content is necessary, but I felt it was sufficiently practical as an auxiliary tool for learning and research.
Moving forward, I would like to try using it for more complex optical design and image processing problems to see how well it can handle them.