Making GenAI Tools User-Centric and Jargon-Free

Today, I'm addressing a disconnect many experience with GenAI tools.

Imagine hiring a builder for office modifications. You detail your needs, but instead of getting started, they ask, "Should I use a hammer, screwdriver, or nail gun? What screws do you prefer?" It's perplexing — you trust experts to choose the right tools.

This confusion is what we face with AI applications, especially those using GPT technology. Users are prompted to select between versions like GPT-3.5 or GPT-4 and adjust settings like 'temperature.' These choices, while meaningful to developers & enthusiasts are just jargon to the average user.

Users want solutions, not technical decisions. They expect technology to navigate toward the solution, not ask them about the subtleties of model iterations or temperature settings. They need AI to understand their goals and deliver coherent, reliable results.

In legal tech, we can tailor tools for tasks. Proofreading a document? GPT-3.5 is suitable, with a lower temperature setting for predictable responses. But for general-purpose interfaces, dynamic adaptation is key. The system should discern the user's goal, letting GPT-4 decide the optimal approach and settings, explained in a JSON-formatted response.

“Given a user's specific task requirements, I need a system recommendation on whether to use GPT-3.5 or GPT-4 for the job. Additionally, advise on the appropriate 'temperature' setting for the chosen model, ranging from 0 to 1. Please provide the response in JSON format, including the rationale behind the model and temperature selection based on the task's demands.”

By adopting a user-oriented stance, we bridge the gap between human expectation and technological delivery, ensuring technology is not just powerful but also accessible and empathetic.