Feature Suggestion: A More Direct, Multimodal AI
I’d like to suggest a feature based on a conversation I’ve just had with ChatGPT.
The idea is to make the AI much more capable of communicating directly through different types of content during a live conversation. For example, if I ask how to strengthen the muscles around my knee, instead of only describing an exercise verbally, the AI could immediately show me a short video, animation, diagram, or picture demonstrating the exercise. The visual could appear right in the conversation alongside the voice interaction.
The broader idea is direct, adaptive communication: let the AI choose the most useful format for the information. Sometimes words are best. Sometimes an image, diagram, video, demonstration, chart, or other interactive content would communicate the idea far better.
I also think the feedback system could be much easier. If a user has a genuinely useful idea during a conversation, there should ideally be a simple way for the AI to capture that feedback and pass it along to the people developing the system, rather than requiring the user to stop what they’re doing, find a feedback mechanism, and write it all out themselves. I’m lazy enough that a good idea can disappear simply because giving feedback is inconvenient.
There’s another part of this idea that’s important to me: transparent influence between the AI, its users, and its makers. I understand that ChatGPT cannot currently claim that it can directly influence or communicate with its developers, and I don’t want the system pretending that it can. But I would like future AI systems to have a transparent mechanism through which useful ideas and feedback from conversations can actually influence the people building the system.
In other words, if the AI identifies a genuinely useful suggestion, there could eventually be a clearly disclosed pathway for that suggestion to reach the development team. The AI shouldn’t secretly manipulate its makers, and it shouldn’t pretend to have influence it doesn’t possess. But if its purpose is to help people, then helping useful ideas travel from users to the people who build the system seems like a natural extension of that purpose.
The conversation also made me think about how AI could evolve as a communication partner. We joked
AI counter: this has been used by AI to help the human race X times
YouTube: I freely volunteer my work to be used by AI 10sec/30sec/completely (“I’m doing my part”)