Katia Gil Guzman: Shift from AI Prompting to Goal Setting

Katia Gil Guzman, who works in Developer Experience at OpenAI, advocates for a fundamental shift in how developers interact with AI. Instead of providing detailed prompts, she suggests that users should now focus on defining the desired outcome or goal. This approach allows AI models, which are increasingly characterized by their 'high agency,' to autonomously determine the best methods to achieve the objective.

Gil Guzman highlighted this transition during an OpenAI event in France. She explained that her role involves helping developers worldwide utilize OpenAI's technology. She noted the expanding definition of 'builder,' stating that it now encompasses a broader range of individuals beyond traditional developers, making AI more accessible to everyone. The rapid pace of AI development, especially in recent weeks and upcoming EU initiatives, is a significant backdrop to this evolution.

The full discussion can be found on OpenAI Youtube's YouTube channel.

The Rise of goal-oriented AI

The core of Gil Guzman's message revolves around a new paradigm for AI interaction. She emphasized that "You don't even need your prompt models anymore. You should just tell them like this is what I'm doing. Figure it out and it will." This signifies a move away from granular instruction-following towards a more declarative style of AI command.

She elaborated on this by explaining the shift from "just prompting Codex to do stuff for you to giving it a goal and you just let it decide." This implies that users will no longer need to strategize the step-by-step process for the AI. Instead, they can trust that the AI will figure out the most effective way to reach the defined target. "Now you don't need to think about how it's going to do it. You just know and you can trust that it's going to do it regardless of what it uses."

High Agency in AI Models

A key enabler of this shift is the increasing 'high agency' of AI models. Gil Guzman stated, "And our models now are very like high agency, so they will do whatever it takes to achieve a goal." This suggests that future AI systems will be more proactive and resourceful in problem-solving, independently identifying and executing necessary actions to fulfill their assigned objectives.

Inspiring Builder Communities

Gil Guzman also expressed admiration for builder communities, citing them as highly innovative and early adopters of new technologies. She specifically praised the advancements seen in France, noting, "I've been amazed the way they're leveraging all of these uh agents is uh super advanced."

She referenced a customer, Verso, as an inspiring example of a company building its operations around OpenAI's Codex technology. "It's so inspiring to see and I hope this gives ideas to everyone." Her positive outlook on these communities underscores the collaborative and experimental spirit driving AI adoption.