Arizona State University students are learning to build personal AI agents that function as extensions of themselves, keeping all data private and under their control. Nearly 80 students across ASU's Tempe and Los Angeles campuses are enrolled in "The Agentic Self," a groundbreaking course taught by will.i.am, the Black Eyed Peas frontman and founder of FYI.AI. These AI agents go beyond traditional chatbots like ChatGPT or Claude by reasoning, researching, and completing tasks autonomously while reflecting each creator's unique values, voice, and goals .

What Makes These AI Agents Different From ChatGPT?

The distinction between agentic AI and conventional large language models (LLMs) is fundamental. Large language models are AI systems trained on vast amounts of text data to predict and generate human-like responses. While ChatGPT and Claude store user data in public cloud systems, the agents students are building through the EDU.FYI platform remain entirely private and personal. Will.i.am explained the core difference: agentic AI represents the next evolutionary step where agents can execute tasks and workflows on your behalf with minimal ongoing input .

Dhruv Patel, a third-year computer science student, created an agent that knows what he knows, shares his perspectives, and even mimics his speech patterns. "I want my agent to be a riffing buddy or a friend I can bounce ideas off of," Patel said. "I want it to think like I do, but with all the computational power and all the other intellectual capabilities that I don't have behind it" . This represents a fundamental shift from asking an AI tool for help to having an AI collaborator that understands your thinking process.

"Agentic is the next step, where the agent is able to do tasks and workflows on your behalf. You set it on its course, and it would reason, research, browse, generate, all autonomously," said will.i.am.

Will.i.am, Professor of Practice at Arizona State University

How to Build Your Own Personal AI Agent?

• Define Your Values and Goals: Before building an agent, students first identify what they believe in, what they care about, and what they would never compromise on. This foundational step ensures the agent reflects authentic personal priorities rather than generic corporate objectives.
• Train the Agent With Your Data: Students input their own materials, perspectives, and work samples into the EDU.FYI platform. Because the data remains private and under student control, creators own everything they contribute without risk of corporate scraping or unauthorized learning.
• Enable Autonomous Task Completion: Once trained, agents can reason through complex problems, research topics independently, and generate solutions with minimal human intervention. Students set the agent's course and let it work autonomously within the parameters they establish.
• Iterate Based on Results: The course teaches students to evaluate how their agents perform and refine them over time. This continuous improvement process ensures agents become increasingly aligned with their creators' thinking patterns and goals.

Why Is Data Privacy a Game-Changer for Artists and Creators?

The privacy-first approach has particular resonance with creative professionals who have grown skeptical of AI after streaming and social media disrupted the entertainment industry. Peter Murrieta, an Emmy-winning producer and interim associate dean of the Herberger Institute for Design and the Arts at ASU, emphasized this concern. "Artists, including myself, are rightly suspicious of AI after the entertainment industry was disrupted by streaming and social media for the profit of corporations," Murrieta noted .

The EDU.FYI platform addresses this anxiety directly. Murrieta explained that the course asks fundamental questions before students build their agents: "What do we believe in, what do we care about? What would we never give up?" The ability to create an agent where you own all materials you input makes it "much easier to move forward in the world," he said. For artists specifically, the guarantee that "anything I tell it, I own, and nobody's scraping and learning off of it" feels like "the only way to move forward as an artist" .

Murrieta

Jeremiah Holland, a graduate student in ASU's Narrative and Emerging Media program, took the course specifically because of this data ownership model. "Right now AI is owned by the big corporations that get to use your data and you don't have any say. With this class and with this AI system, the data is under our control and we get to decide how it's used," Holland explained . This represents a philosophical departure from how most AI tools operate today.

How Does This Fit Into ASU's Broader AI Strategy?

The "Agentic Self" course is part of a larger institutional commitment to AI education at Arizona State University. In 2024, ASU became the first university to partner with OpenAI. The partnership expanded last year to bring ChatGPT Edu with GPT-5 to every student, faculty member, researcher, and staff member at no cost to individuals . This means ASU students have access to OpenAI's most advanced models while simultaneously learning to build private, personal agents through will.i.am's course.

The university offers two undergraduate and 10 graduate degrees focusing on AI, including a one-of-a-kind degree in artificial intelligence engineering. Will.i.am calls this comprehensive approach "forever learning," a way to equip students in any major with skills they need for an AI-powered world. He emphasized the moral dimension of this work: "It's a moral compass, this urgency to teach, inspire, encourage, mentor and motivate folks to build ethical systems that reflect themselves" .

The partnership between ASU and will.i.am's FYI.AI began when ASU President Michael Crow pitched the idea directly. "President Crow pitched, 'Hey, those agents that you're building at FYI? What about building them for professors and students?' That vision is the reason why I'm here," will.i.am said . The EDU.FYI platform being tested in this course will eventually be available to all ASU faculty and students, with plans to expand to other educational institutions.

What Practical Skills Are Students Actually Learning?

Beyond the conceptual framework of agentic AI, students are gaining hands-on experience with real-world applications. Patel, who grew up listening to the Black Eyed Peas, said he registered immediately when the class opened. "It's one thing to learn agentic AI from a computer science professor talking about all the components or the hardware side of things. It's another thing to learn it from this creative genius who can explain it to me in a way that only an artist could," he noted .

Holland has already used other AI chatbots like ChatGPT and Gemini for practical tasks such as fact-checking and grammar review. When writing scripts, he uses AI to generate character name suggestions and brainstorm ideas. But building his own agent from the ground up offers something fundamentally different. "This is the first time I get to build it from the ground up and decide how it thinks and how I want it to help me. Being able to see the back end of how these systems are built is one reason I'm glad I took this course," Holland said .

When

The course structure itself reflects will.i.am's commitment to accessibility. He alternates between teaching in Tempe and Los Angeles, so each week, half the students see him in person while the other half watch via video stream. This hybrid model ensures that students across both locations receive direct instruction from a figure who bridges the worlds of entertainment, technology, and education.

As AI becomes increasingly central to how work gets done, the question of who controls AI systems and how they reflect human values becomes more urgent. The "Agentic Self" course suggests one answer: students who understand how to build agents that think like them, respect their data, and amplify their capabilities without compromising their autonomy.