AI is the new UI

AI is the new UI
I view computing as one of the most valuable resources for the betterment of life on Earth that mankind has discovered. Since I began my career, that resource has never been fully tapped. The value of computing to mankind has been constrained by our ability to use it effectively. That's why I became a software developer, and why we named our company GenUI: because we believed great value could come from working to make computing easy for everyone to use. Recently and suddenly, it's become apparent to me that computing is about to get much, much better for everyone.

The Democratization of Computing

The history of computing has seen three major revolutions that have expanded access to computing and extended its benefits to more and more people. In the 1980s, the personal computer took computing out of government and academic labs and into the homes of millions. Starting in the late 1990s, the internet took off, transforming the world and empowering people across the globe with information and services. The third such revolution, in the late 2000s, was mobile computing - empowering billions in every corner of the world with apps and connectivity, and now over 90% of the global population owns or has access to a mobile device. 

Today, the fourth major transformation in the accessibility and use of computing is underway, taking the form of LLM-powered AI agents that will reduce the end-user burden inherent in today's technology products. We're on the threshold of an age in which the role of the user may be played by an artificially intelligent agent instead of a human, and this substitution will represent an extraordinary acceleration of humanity. And, importantly, this will bring the full power of computing to everyone.

The Burden of User Interface

Technology brethren like myself have generally been willing to devote the mental energy to learning how to set up, configure, and put to use all the features of the panoply of smart devices and software products that surround us. But I will admit that lately it feels burdensome to delve into all the features of a smart TV, or even to comb through the Settings menu of the latest iPhone. The truth is that most people don't care and can't be bothered.

The smart home of today has become a wasteland of forgotten or unused bells and whistles that served to sell the product but provide little value after the fact. As a UI developer or designer, there's enormous compulsion to design for ultimate simplicity to avoid confusing or repelling an uninvested user. Designing for the lowest common denominator in a kind of "race to the bottom" of features and functionality has massively slowed the adoption and value of technology products over the whole history of the industry. But what would be possible if we didn't have to sacrifice functionality on the altar of user experience?

Agents as Users

This year's advancements in generative AI are inspiring, because it is now not only possible but relatively easy to create an AI agent that can operate technology products on behalf of the human user. Developers are actively building ChatGPT Plugins that allow the GPT-4 model to understand, interact with, and control the interfaces of other software products and smart devices. These AI agents can do the heavy lifting to understand how a product works before tailoring its setup, configuration, and operation to the human's specific needs. When this happens, our technology products will feel magical: personalized, frictionless, automatic and so intelligent they seem sentient.

Those of us who design and build technology products now have an entirely new paradigm. Products can now be designed for AI agents to use on our behalf, rather than for humans to use directly! In the past, a smart washing machine may have had a touchscreen or an array of buttons, but in the future, it will have nothing but a lid. A task like setting up a Wifi router - intimidating for many in the past - may become as easy as asking an agent to get it done.

And, excitingly, the interface (or virtual "control panel") we design for an LLM agent to use can be arbitrarily complex. No longer will we have to reduce the feature set of a device's UI or dumb it down for the benefit of a human user. If a feature exists that will serve the human's needs, the agent will use it (optionally letting the human know what it is up to). The kid gloves are off.

It's not just consumer products, by the way. This same change is coming to the way businesses interact with one another, how we use software platforms and development tools, and especially to the way to we visualize and interact with data.

New Challenges for Builders

Product visionaries have just been handed lightning in a bottle. There's never been a better time to bring a new value proposition to market, or to modernize an existing enterprise. But to be sure, there's a lot of work to do. One of the hallmarks of at least the current generation of LLM-based agents is that while it's trivial to create an amazing demo, a lot more work and care is required to create a robust product. The agent must be supplied with complete information about available features and functionality, but also with personalized information about the user and situational context. We've got to put good guard rails up to ensure that the agent is trustworthy, and that its operation will be secure, compliant, and privacy is preserved. And we may push intelligence to the edge instead of the massive centralization that exists today.

Generative AI is about to once again drastically democratize technology by reducing friction: simultaneously lowering the barriers of use and increasing the benefits derived. Technology will be less encumbered by the readily apparent downsides of today, ranging from excessive screen time to frustration and confusion. I hope, and believe, that everyone on Earth stands to benefit greatly.

AI is the new UI.