[Hugging Face] - Artificial Intelligence Agents On The Move! *Tech Take*
The Technical Take:
Hugging Face, which is an artificial intelligence company that provides open-source models to engineers, has released a course on artificial intelligence agents. According to Hugging Face, artificial intelligence agents are systems that leverage AI models to interact with its environment in order to achieve a user-defined objective. They combine reasoning, planning, and the execution of actions (often via external tools) to fulfill tasks; at a high level, there are main two parts of an AI agent: the Brain, which is commonly Large Language Models (LLMs) that think and plan, and the Body, which are the tools and capabilities that perform the action(s). For example, if I want an AI agent to retrieve the weather and forecast for the week, I would program the AI agent to have the capability through function calls and correct formatting; particularly, an Application Programming Interface (API) call would be made to a weather application or website, returning a Java Script Object Notation (JSON) format that has the weather and location, like New York. An example of a code agent’s retrieving weather information:
Code snippet from Hugging Face
Concisely, the code snippet defines a function that retrieves weather of a particular location through an API call. If the request is 200, which is successful, return the weather (data = response.json ()) in JSON format that is parsed into a Python dictionary. The data that is returned either has the weather, no weather information available, or there has been error in which the weather cannot be retrieved; the AI agent can be used singularly to find weather, or it can be layered with other tools to retrieve information from the internet; nevertheless, I find the topic of AI agents interesting because there is a misconception that artificial intelligence chatbots are autonomous and can think for themselves, yet that notion is incorrect. Engineers have to code these tools, actions, and observations into the agent for it to appear autonomous; the reinforcement learning mechanism, as well, is coded into the agent to adapt its search strategy to provide better results. Just as these features can be coded into the model, it can be removed and become a simple chatbot. As with all AI technologies, it can used and infused with data that is representative of its users and free of maliciousness. Hugging Face has given developers the ability to create models that are representative of its users, which are models with their own stories of creation; those who are underrepresented, such as Native Americans, can create models that preserve their language and other important aspects of their culture. I believe in their vision of open-sourceness, and I am glad to be taking a class that would enable me and those who look like me to tell our story through code.
I have already completed Unit 1 of the course, and I created, modified, and examined my first AI agent. I basically duplicated the source code from the course, and I began creating tools that AI agent can use to help me in researching things, like retrieving the prices of the Apple products, creating images from text messages, and finding the president based on the calendar year. I believe I have a good handle of AI agents, and I plan to create more advanced ones to refine my skillset in machine learning. Until then, I have a task to complete that is to complete the course and use what I will have learned for work and personal projects.
The Reference:
“Welcome to the 🤗 Ai Agents Course - Hugging Face Agents Course.” Welcome to the 🤗 AI Agents Course - Hugging Face, Hugging Face, huggingface.co/learn/agents-course/en/unit0/introduction. Accessed 22 June 2025.