Artificial intelligence is moving fast. The way people interact with it will evolve as the functionality improves. These are my predictions of where the Human-Computer Interaction (HCI) is going,
Pre- ChatGPT
It used to be called ML or machine learning. I worked at a startup a few years back that used a very advanced system to monitor trillions of signals from huge things like aircraft carriers or steel mills. The system would figure out how what parts needed servicing before they even broke based on combinations of many signals. The UI was all about training and giving feedback, plus a boatload of dashboards.
Additionally, we had Alexa, Siri, and Hey Google. All of these could take basic questions and answers. Alexa is especially good with setting alarms using voice commands.
In this era, natural language was limited and only used in short bursts.
ChatBot UI
ChatGPT is the big meteor that shook the world. The interface was basically taken from the patterns of texting. You asked, it answered. It seemed almost human and had lots of information. Unfortunately, these systems had very poor memory and would frequently hallucinate.
Chat interfaces are actually pretty bad. Typing what you want in full sentences sounds good in theory, but in practice it’s painfully slow and error prone. The hallucinations and outright lying by the chatbots made it worse and diminished trust
Vibe UI
There are a slew of new interfaces that integrate the natural language input with different modes of output. I saw a demo of SubFrame, that let you design a user interface with AI integrated into every layer. Claude, MS CoPilot, Google Gemini and others and other programming centric tools are being integrated into common coding tools like VS Code. The term for using AI in this way is vibe coding or vibe designing.
The actual capabilities of vibing are still in flux and changing on a weekly basis.
This interaction pattern is compelling because you can use your normal output, but with a helper that can do repetitive tasks quickly. (If it doesn’t break every minute!)
Agents
The Agentic UI is being discussed heavily right now. Basically, it is a specialized AI that can act on your behalf. You can for example say that you want to travel to New Zealand. It will buy the airline tickets, rent the AirBnb, and get you a car all by itself.
In business circles, Agents can be set up to all manner of tasks. There is a technological framework called MCP, that helps companies connect their AI to different services.
The UI of how to set up an agent, monitor and manage it are all in their infancy. I think this can not be just a chatbot. You need to have ways to review options and approve what the agent is doing.
The Era of Validation UI
There will come a day, not too far into the future where a major portion of our user interfaces will be to validate the actions of our agents. It may be similar to an inbox if you have many agents doing things all at once.
I believe we will need to invent new gestures to handle the load. The closest thing I can see right now is the Slack mobile “Catch Up” interface.
You have to use it to fully understand the paradigm, but basically it is either acknowledging the agents actions (or not) or talking to it about the action before approving. I think this is the UI that will be on many of our systems.
The invisible UI
There is an old saying, “I don’t care how good your UI is — less of it would be better.”
After a while, we will see that we are rubber stamping the UI approvals on many items. We will start to trust the system. At this stage, we will want to say “Just auto-approve it”. This is when AI agents really start to shine because they will be protecting and servicing us without us even knowing it.
Food will just show up on our door steps. Tickets to shows, dates with single people, dinner parties, trips, job offers, everything.
Everything will just happen. We will not even remember after a while how it’s happening. We will tweak preferences over time, but the AI memory and trust issues will have been long forgotten.
People will not know how it’s all happening, but it will. It’s the way I feel I about movies. I don’t know who makes them, but I like watching them and they just show up on Netflix without me doing a thing.
What then?
There is a complex question about capitalism that this raises. What jobs would people even have at this point? Is there a world where everything is free? What does that mean for the human condition?
This post isn’t about that, so you will have to just wait until I write about that.
Let’s just say, I am optimistic that we will have a happy ending.
Whatya think?