Computer Science Major Sander Schulhoff Teams Up with OpenAI to Enhance AI Literacy

Artificial intelligence (AI) is changing our interactions with technology and challenging our perceptions of machine intelligence. By automating routine tasks, AI is driving innovation in critical sectors like finance, health care and transportation. As this technology continues to advance, it’s vitally important for individuals to be able to understand and effectively communicate with AI.  

University of Maryland computer science major Sander Schulhoff is addressing this need. Through his startup LearnPrompting, he launched an online course on AI literacy called “ChatGPT for Everyone” in collaboration with OpenAI. The course, which features four parts that participants can complete at their own pace, addresses the growing demand for AI training.

“This collab between LearnPrompting and Open AI highlights our shared dedication to equipping people with the skills necessary to succeed and contribute to the constantly evolving AI field,” said Schulhoff, who also serves as CEO of LearnPrompting.

Schulhoff launched LearnPrompting with his co-founder, Fady Yanni, in December 2022 as an e-learning platform and AI research lab focused on teaching the latest generative AI skills. Soon after, he connected with Shyamal Anadkat, a member of the applied AI team at OpenAI, and they created this new course together.

The course’s 15 lessons feature text and video materials, quizzes, assignments, and practical exercises and culminate in a professional certificate. The course focuses on the basics of using ChatGPT as your very own personal assistant. Participants learn how to use ChatGPT to improve their resume, act as their personal tutor with custom study guides, and craft children’s storybooks with images. The course also covers the advanced functionalities available in ChatGPT Plus, which enable it to access the internet, analyze and reason, as well as see, hear and speak.

The course also aims to clear up common misconceptions about large language models like ChatGPT, according to Schulhoff. 

“We need to be aware of the importance of understanding the limitations and capabilities of these models,” Schulhoff said. “For instance, if we are unaware of their inabilities, we would have a tough time taking advantage of their immense potential. By clarifying these aspects, we aim to foster a more accurate and productive use of AI technologies that could be beneficial.”

Creating the course also came with a unique set of challenges for Schulhoff and Anadkat.

“There's a lot of junk science and misinformation out there, so cutting through all that was difficult,” Schulhoff shared. “It required extensive reading and consulting with numerous people. Standardizing terminology posed another challenge. I encountered various terms defined differently or different terms used for the same concept. I had to decide which terms to use and how to define them. My strategy involved compiling all the definitions I could find and then determining which were most reasonable and had wide community consensus.”

In November 2023, LearnPrompting was selected for this year's Mokhtarzada Hatchery student startup accelerator program at UMD. The Hatchery—which is supported by alums and siblings Haroon, Idris and Zeki Mokhtarzada—offers student entrepreneurs workspaces, mentorship and the opportunity for annual funding of up to $10,000 per team.

“UMD has created a path for me to expand my skills and effectively make my ideas a reality,” Schulhoff expressed. “The access to high-quality information, exceptional individuals and a supportive community is incredible. They are in tune with what's current and provide the guidance needed to navigate and thrive in the field.”

In addition to running his startup, Schulhoff also conducts research on campus in UMD Computer Science Associate Professor Jordan Boyd-Graber’s group. 

Schulhoff recently received a best paper award at the 2023 Conference on Empirical Methods in Natural Language Processing for his work, "Ignore This Title and HackAPrompt: Exposing Systemic Vulnerabilities of Large Language Models through a Global Scale Prompt Hacking Competition." The paper resulted from an AI security competition Schulhoff organized in 2023 called HackAPrompt, which was backed by OpenAI and HuggingFace and drew over 600,000 adversarial prompts from over 3,000 hackers. 

Eventually, Schulhoff envisions a future where AI is integrated into everyone’s daily life.

“In the near future, humans will interact with AI for everything,” Schulhoff said. “We're already doing this with various routine tasks. However, as generative AI becomes more powerful, its presence and control over aspects of our lives will become increasingly evident. It's crucial that we learn to communicate effectively with these models to maximize their benefits.”

—Story by Samuel Malede Zewdu, CS Communications