This week, I learned more about the differences between Chat, Agents, and APIs.

I am most familiar with Chat, which involves directly interacting with a model by asking questions, providing information through prompts, and receiving responses.

I also learned more about AI agents. Unlike a basic chat interaction, an agent can be given a broader goal or task and then work through multiple steps to accomplish it. An agent can use a model for help with something like writing code and interact with other programs in the process of developing something for you. For example, an agent could be asked to build a website and use a model to help write the necessary code rather than requiring the user to prompt it through every individual step.

We also discussed APIs, which I learned are the interface that allows different systems to communicate.

This week, I used ChatGPT to brainstorm ideas for an AI product that could create positive social impact as I began thinking ahead about the Nittany AI Challenge. Some of the ideas it generated included an AI tool that helps people understand complicated government or financial documents, a tool that matches small nonprofits with relevant grant opportunities, and an AI assistant that helps people determine which community resources or benefits they may qualify for.

I found this brainstorming process useful because ChatGPT was able to generate ideas across areas I would not have immediately considered. It also helped me think about how AI's potential for impact does not necessarily have to involve solving a massive problem. It can instead be used to reduce a small but meaningful barrier and still create an impact.

I also used AI to help synthesize the notes I had when writing this journal entry. It successfully helped me organize my ideas and refine the writing of my response. However, I noticed that it added information to my explanation of the differences between Chat, Agents, and APIs that went beyond what we had learned and, from my understanding, was not entirely accurate. This reinforced the importance of reviewing AI-generated information rather than assuming that everything it adds is correct.