Is Personalized AI the Ultimate User Agent?
For those of you familiar with web development, the user agent is a common concept. It’s the way that a piece of software communicates identifies and communicates itself to other machines on the network – so, if you’re using Firefox to access Google to search for something, your computer is sending a little bit of information to Google to say something like: “Hey, I’m Firefox, and I’m running on a Linux machine using this version of Firefox”.
If you’ve ever seen a popup that says something like “This feature is only available in Chrome/Edge”, you’ve seen the issues with anti-competitive web browser monopolies… and you’ve observed the user agent in action – that’s how they know you’re using a different browser from the one that they want you to be using.
The Mozilla Developer Network describes a user agent as:
A computer program representing a person, you say? Where have we seen that recently? Oh right – everywhere.
The ultimate potential for artificial intelligence isn’t – despite what Microsoft, Google, and Open AI want you to believe – a large language model trained on the entirety of the world wide web. The far more exciting potential is in what smaller, local models that are highly customized to an individual user and task will be able to do. We’re about to witness a glorious return to “home grown” software, where smaller communities and organizations are able to utilize open source technologies to create personalized interfaces to an increasingly accessible amount of data, information, and other people.
The role of the browser* is going to change. How much – and how quickly – is TBD: rapidly changing technological and regulatory environments make it hard to predict market trends. But one thing is for certain: we need to change our mental models of the internet. I don’t care if you want to call it web3, web3.0, web5 (web 2 + 3?), metaverse, fediverse, or Bob. What matters is that we create new frameworks and tools for navigating the amount of shared thought and information that we’re now exposed to on a daily basis.
I realistically, today, already have many user agents that represent me, albeit it doesn’t carry quite the same weight when I’m just communicating within my private communities. The line between digital agent and my self-described identity is blurring; I have accounts and programs that represent me, owned by companies like Meta and Google and Mozilla and Microsoft and Columbia University and… the list goes on.
With the rapid increase in capabilities of machine learning technologies, though, I’m optimistic that the world of open source (and computer science) is going to become more comfortable for people to navigate, so that they can truly take back agency with how they operate with these digital agents. I’m not saying that there’s no world for commercial software, but I think trends in decentralization point to a growing amount of discontent with the status quo.
The demand for trustworthy user agents will only increase in importance as identity becomes easier to fake. We need to be considering paradigms where we are granting actual user agency to our user agents, which isn’t to say that the role of the browser or browser provider will go away, but it may shift to encompass a wider range of perspectives that influence the way that we interface with the world of ones and zeros.
Do you trust your user agent?
This latest wave of attention on machine learning technologies is forcing us to yet again consider the position that wealthy technology companies are in. Bing, Bard, ChatGPT, and now IBM’s ‘Dromedary’ AI chat bots are all injecting a western, corporate, mass-appealed censorship layer into their language models to comply with regulation. The balance between safety and openness is a tightrope, which we’re seeing play out in real-time.
I’m betting on the local models. I dream of a day where the future of our operating systems are deeply personalized, and owned, as our own – when we are the authors of our own user agents.
* Disclaimer: I work for Mozilla, but not on Firefox. This is not a statement that reflects company opinion, but is my theory based on my personal philosophy of how we build and engage with technology.