Back in January, we tried using Chat GBT to solve integration problems and concluded that Chat GBT is able to perform integration fairly well. However, in recent months, there have been complaints from users that the developers have made changes to Chat GBT, meaning it is not able to perform as well as it used to. Furthermore, a Stanford researcher has written an article concluding that the current Chat GBT version has performed worse in some tasks, particularly in solving math problems.
Today, we are revisiting this topic to evaluate whether the current version of Chat GBT performs well in solving integration problems. We will be asking the same three integration questions to the newly updated Chat GBT and compare the results to the ones we obtained a few months ago.
The first question we asked Chat GBT to solve is the integration of PSI X. As we can see in the results shown on the screen, the current Chat GBT version provides the same answer as the old version released six months ago. However, one significant difference is that the current Chat GBT provides a more succinct answer than the old version. In the old version, three solutions are provided for the problem, whereas in the current version, we see one answer and a brief explanation of what each individual term means in the solution. This example shows that the ability of the current Chat GBT version to provide a concise solution proves that it is an improvement to the first version.
The second question we asked Chat GBT is to solve the integration of e to the power of x cosine 2X. We compared the two results by looking at the solution steps and concluded that they are indeed quite different. In fact, the current Chat GBT version provides the correct solution to this particular problem. Additionally, it shows the proper way of performing the integration by parts and simplifying the solution to the model answer. This is very different from the old version, which expressed the final answer in a way that is unfamiliar to how we usually do it while solving math problems. Based on this evidence, we conclude that the current Chat GBT version is better because of its ability to closely imitate how to solve math integration problems.
The last question we asked Chat GBT to solve is the integration of one over long X. The solution provided by Chat GBT shows that it is able to identify that this is a non-elementary integral and suggests expressing it as a special function called a logarithmic integral function, which is one approach to solve the problem. The old Chat GBT, however, provides a wrong approach to solve this problem. From this example alone, we can conclude that the current Chat GBT version is one level above the old version in terms of the ability to solve difficult integration problems.
In summary, the new Chat GBT is a vast improvement compared to the old Chat GBT that was released six months ago. If you found this article helpful, please consider subscribing for more interesting Chat GBT AI content. Thank you for reading!
