Universities should arm students with AI ‘eval’ powers

Universities are restricting AI, but a better approach involves teaching students how to use AI’s expertise and evaluate its output through a “loop” process. This method, tested at Stanford, allows students to build critical skills and hold AI models accountable, fostering digital sovereignty.

The “loop” enables organizations and individuals to combine their expertise with AI, test models against their specific goals, and repeatedly optimize them, ensuring they retain control and value.

This practice ensures that students graduate with the ability to measure and command AI systems, preventing a future where people passively consume AI output dictated by corporations.

• Young people feel they have little say in the concentration of power within AI companies.
• Universities are restricting AI in classrooms, hindering students’ ability to understand and wield the technology.
• A better response is teaching students to use AI through a “loop” process of combining expertise, testing AI, and optimizing models.
• This “loop” offers a new kind of sovereignty, allowing organizations and individuals to stay in charge of their destiny in an AI world.
• Teaching this “loop” in classrooms builds real skills, such as systematic testing of model output, and holds AI models accountable.
• Students can create “evals” to measure AI model performance against their personal criteria, demonstrating their understanding and passions.
• The ability to design an eval requires existing expertise in a subject matter.
• The goal is for every student to leave university having built at least one eval of AI systems.
• The divide in the AI age is between those who can measure and bend AI to their purposes and those who cannot.
• Knowledge empowers people to be their own governors, and the “loop” is how individuals arm themselves with AI knowledge.
  Continue reading https://www.ft.com/content/ac15aa4a-960f-4957-90c6-ff1b4a390fa4?syn-25a6b1a6=1