Following the successful Knowledge Sharing Event, the project has reached its conclusion. The team has accomplished a series of significant milestones, including the production of an extensive guidelines report and the development of two Proof of Concept (PoC) applications that vividly demonstrate the potential of generative AI.

One noteworthy aspect we'd like to emphasize is how readily these applications were developed. Despite being conceptual in nature, they effectively showcased the capabilities of generative AI. This ease of development was both striking and pleasantly surprising, as it highlights the rapid progress and accessibility of AI technologies for creating impactful tools, even within the boundaries of proof of concept.

The team takes great satisfaction and pride in the accomplishments achieved over the summer. We are immensely grateful to AI Sweden and Chalmers for granting us the opportunity to contribute to the advancement of AI awareness and applications within education.

Summary (24/07-27/08):

Over the past several weeks, the team has dedicated its efforts to extensive research, writing, and refinement of the guidelines report. This report has been the focal point of our endeavors, aiming to provide a comprehensive and insightful overview of the capabilities and limitations of Large Language Models (LLMs), particularly focusing on ChatGPT. The report can be viewed via the provided link or by viewing the available resources on the project page.

Through rigorous investigation, we have delved into the vast landscape of LLMs, assessing their potential applications and recognizing their boundaries across various domains of education. Our exploration has resulted in a clear understanding of where these models excel and where they may fall short.

The report not only addresses the present capabilities of LLMs but also delves into the potential trajectory of their evolution. Our message emphasizes the importance of adapting educational practices to harness the benefits of these AI technologies responsibly and effectively.

In addition to the guidelines report, the team has created a presentation that will be a fundamental part of the upcoming knowledge sharing event scheduled for August 29th. This presentation aims to disseminate our findings, insights, and recommendations to a broader audience, fostering awareness and promoting responsible AI adoption in the education sector.

The culmination of weeks of research and analysis positions the team to contribute valuable insights to the conversation surrounding the integration of LLMs, paving the way for informed decisions and responsible AI use in education.

Weekly Summary (17/07-23/07)

In the fifth week of the project, the team reached a realization that implementing the "AI Thoughts" concept might exceed the current project's scope. As a result, it was acknowledged that the textbook companion app would primarily remain a proof of concept. Despite this shift, the team continued to fine-tune the app's behavior to align more closely with the approach of a human tutor.

One notable enhancement was the successful implementation of a feature that enables the application to accurately format mathematical expressions. This improvement is demonstrated in the provided image, showcasing the app's ability to handle and present mathematical content appropriately.

Looking ahead, the upcoming weeks will see a shift in focus towards the production of a comprehensive report. This report will provide valuable guidelines on how universities and educational institutions can effectively engage with and adhere to Large Language Models (LLMs) like ChatGPT.

Weekly Summary (10/07-16/07):

During the fourth week of the project, significant progress was made in finalizing the appearance of the Streamlit pages for both the textbook companion app and the code correcting tool. The team focused on refining the user interface and ensuring a visually appealing and user-friendly experience for students and teachers.

To enhance the problem-solving capabilities of the textbook companion app, the team explored the concept of "AI thoughts," which involves enabling the chatbot to think independently before generating a response. This approach aims to make the bot more intelligent and capable of providing more insightful and contextually relevant answers. However, the team is still deliberating on the best way to implement this feature and further discussions are ongoing.

As shown in the accompanying images, the Streamlit pages of the textbook companion and code correcting tool have evolved significantly, offering a more polished and professional appearance. These improvements contribute to enhancing the overall user experience and usability of the applications.

In the upcoming week, the team plans to focus on further refining the problem-solving capabilities of the textbook companion app and continuing discussions on how to effectively implement the "AI thoughts" concept. These developments aim to advance the functionality and intelligence of the app, providing students with more comprehensive and insightful support in their learning journey.

Weekly Summary (03/07-09/07):
During the third week of the project, the team made significant progress in the development of the textbook companion app. To enhance its presentation and user experience, we began exploring the integration of Streamlit, a popular Python library for building interactive web applications. This step aimed to make the app more user-friendly and visually appealing for students.

Furthermore, we delved into the idea of utilizing ChatGPT as a code corrector. With this application, our objective is to create a tool that can assist teachers in grading student code, thereby expediting the grading process and allowing teachers to allocate more time to teaching. This potential use case has the potential to streamline the assessment process and alleviate the burden of manual code grading.

Parallel to these developments, a significant portion of our time was dedicated to extensive research on the topic of integrating chatbots in education. We focused on exploring best practices and formulating a comprehensive set of guidelines for effectively incorporating chatbot technology into educational settings.

Looking ahead to the next week, our goal is to further refine and finalize both the textbook companion app and the code corrector tool.

Weekly Summary (26/06-02/07):

In the second week of the project, our focus shifted towards the development of a textbook companion app, inspired by the transformative impact of one-to-one tutoring on student learning. We aimed to create a platform where every student could have free access to a personalized tutor specialized for each textbook they are studying. Our immediate goal was to create a proof of concept for this app.

Additionally, we continued our exploration of ChatGPT's coding capabilities, particularly in the context of MATLAB. We conducted further problem-solving tests to assess how well ChatGPT could generate MATLAB code for various tasks. This allowed us to gain insights into the model's ability to handle different types of problems and further refine its coding skills.

By focusing on the development of the textbook companion app and continuing our testing and exploration of ChatGPT's MATLAB coding capabilities, we aimed to advance our understanding of how this AI model can be harnessed to enhance educational experiences and provide personalized learning support to students.

Weekly Summary (19/06-25/06):

During the first week of the project, our main focus was on evaluating the capabilities of ChatGPT in generating code for mechanical problems. We also explored how ChatGPT's code generation differed across different programming languages.

To assess ChatGPT's abilities, we conducted tests using Python to simulate, solve, and derive the equations of motion for a double compound pendulum. With a single prompt, ChatGPT successfully accomplished this task for a double mathematical pendulum. By further fine-tuning the prompts, ChatGPT solved the task without encountering any issues.

Additionally, we attempted to utilize ChatGPT to generate MATLAB code for simulating a specific "stacking blocks" problem, formulated similarly to an exam question. However, ChatGPT faced more challenges in solving this problem compared to the Python counterpart.

The discrepancy in performance likely stems from the differences in generality between MATLAB and Python, as well as the more specific nature of the physics problem itself. MATLAB is not as widely used as Python, and the problem's context may have presented additional complexities for ChatGPT.

Overall, the first week of the project involved exploring ChatGPT's code generation capabilities for mechanical problems and examining the variations observed across different programming languages