Why isn't ChatGPT used more extensively?

Many of you might be familiar with ChatGPT. It's a revolutionary tool that has the potential to transform how we interact with technology, offering capabilities that range from answering questions and generating creative content to providing in-depth explanations on a wide array of topics. Yet, despite its impressive capabilities and the buzz surrounding it, it seems like only a few people are truly leveraging its full potential on a regular basis. So, why isn't it being used more extensively?

In this article, I will delve deeper into why giving instructions to ChatGPT is particularly difficult in a work context, explore the limitations of its current chat-based interface, and discuss potential solutions that could make AI tools like ChatGPT more accessible and useful for a broader audience.

Freedom paradox

ChatGPT offers an incredible amount of freedom. It can write essays, answer questions, generate ideas, and even engage in creative writing. However, this very flexibility can be overwhelming. Imagine having a Swiss Army knife with endless tools but no clear manual on how to use each one—it's easy to get lost.

Similar to management

ChatGPT is like an extremely capable new employee who knows a lot but nothing specific about your job or company. This makes clear communication essential. If you provide precise instructions, you get excellent results. But vague directions can lead to confusion and wasted effort.

Moreover, ChatGPT operates in a chat format, forgetting previous conversations as soon as a new chat starts. This means you constantly have to give detailed instructions, as if you're always dealing with a new hire. Although the new "Memories" feature helps, it’s not a complete fix.

Using ChatGPT effectively requires you to be clear and specific, much like managing a team. You need to outline the task's background, purpose, and resources, which can be cognitively demanding.

Curse of the chat UI

The chat interface of ChatGPT is both its strength and its weakness. It allows for natural conversation but requires users to articulate precisely what they want and how they want it done. This makes it primarily useful for the top 5-10% of people engaged in intellectual activities. Even for them, it's not always user-friendly.

To mitigate the complexity of inputs, OpenAI has introduced features like GPTs, where users can use well-crafted prompts by third parties. However, these features still primarily benefit those who were already skilled at using ChatGPT, enabling them to use it even more effectively.

Most people struggle from the get-go, either when initiating a chat or opening the app. Because ChatGPT operates as a chat UI, users must constantly articulate "what" they want and "how" they want it done.

Object-Oriented UI advantage

In contrast, many apps adhere to object-oriented UI design, where users can start from a list of objects, make selections, and proceed without much thought. These designs reduce friction and make actions simple and intuitive.

Consider the simplicity of using an app like UberEats. When you open the app, you’re immediately presented with a list of restaurants (objects). You don’t need to type out your entire request. Instead, you browse, select a restaurant, and make your action like view details, make reservations, or write reviews without much thought. Each step is guided and intuitive, significantly reducing the cognitive load on the user.

There's a growing recognition of the need for more specialized, purpose-driven interfaces that reduce friction and allow actions with minimal effort. These interfaces could bridge the gap between the flexibility of a chat-based AI and the ease-of-use of object-oriented design.

For example, in project management tools, integrating AI in an object-oriented manner could involve users interacting with tasks and milestones directly. Users might drag a task to an AI assistant icon to get suggestions or updates, streamlining the process and making AI assistance a natural part of the workflow.

Promise of AI agents

AI agents, where the AI makes necessary decisions and actions on its own, are being developed by many tech companies. For example, in a work setting, an autonomous agent could understand internal documents and meeting content to provide the necessary information without requiring detailed instructions.

Unlike ChatGPT, these agents would need minimal context for instructions and would independently seek out necessary information. If ChatGPT is like an excellent new employee, an autonomous agent would be like an excellent mid-level employee who picks up tasks on their own, gathers the required information, and completes the work.

When superior autonomous agents are provided through interfaces beyond just chat, generative AI could reach even more users. Currently, we are in a phase of trial and error, exploring new ways to utilize generative AI beyond chat.

In conlusion, while ChatGPT is a powerful tool, its full potential is not being realized due to the challenges of providing clear instructions and the limitations of the chat UI. The future lies in developing more intuitive AI interfaces and autonomous agents that can operate with minimal user input, making advanced AI accessible to a broader audience. Imagine a world where AI anticipates your needs and acts proactively—it's closer than you think, and it will revolutionize how we interact with technology.