Software Agents

Software Agents

There's a four-book science fiction series called Otherland that was published back in 1996-2001 which presents some interesting takes on what we might today call the metaverse; I haven't revisited the series recently (so I don't know how well it holds up, today), but I remember the world-building as being especially solid (the books are brick-like tomes, and a lot of those pages were spent fleshing out the late-21st century world in which the plot plays out).

One innocuous, but to me quite compelling element of this storyline is what's referred to in the book as "gear," but which we might also call a "software agent."

This agent takes the shape of a cartoonish character called Beezle Bug: a program that interacts with one of the main characters, and which that character has owned since he was younger—hence the cartoonish visage. 

But the idea is this software agent—a program that works on behalf of its owner, a bit like a very smart, personalized Siri or Alexa voice assistant—does all sorts of things for its human companion. And if you own one for a long time, it will learn and change and grow with you, becoming more specialized and more valuable as it adjusts and evolves and figures out how to do more stuff.

This evolution comes through upgrades to the technology, but it also comes from exposure to a single human, which means while this technology might begin as a single, one-size-fits-all product (again, like today's voice assistants) over time it becomes custom-tailored based on its interactions with its owner.

A bit like an imaginary friend who you can engage with via a little robot in real life, but who also exists in the virtual world, and can play games with you, help you do your homework, help you do research into something you're keen to know more about, and maybe even help you hack systems or do illegal stuff, if properly evolved and iterated.

This isn't a concept unique to Otherland: software agents, sometimes called AI (artificial intelligence) agents are a common trope within fiction, and within actual, real-world research and developmental spheres.

But it's something that I've been keeping my eye on as I suspect such agents, as they become more versatile, might seriously augment the capabilities of a learning-focused, multidisciplinary person, both in the sense that they could help surface new things that might be relevant to their owner's interests, but also in the sense that they could be like a virtual Iron Man suit—not increasing strength and granting laser weapons, but increasing cognitive heft and providing all sorts of self-educational tools to those who use them and learn how to wield them optimally.

Proto-versions of these agents exist today, but primarily in the voice assistant space and in the world of non-personality-having tools (many complex tasks are being automated in a "smart" way, so everything from photo filters that can intelligently remove or add things to the background of your vacation snapshots, to audio playback that adjusts to the atmospheric noise in your environment, to search tools that "learn" what sorts of things you typically like to search for and auto-adjust your results accordingly might fit within this category).

My personal ideal version of this software is similar to what was in those books I read when I was younger. I'd love to have such a tool with me at all times, I'd love to be able to engage with it while doing other things, almost like talking to myself but to have it then work on my behalf, completing tasks I give it but also working on things I wouldn't necessarily think to do or have the capacity to do as a non-software human.

In practice, this would likely involve a lot of filtering and searching for information and media, wading through all the writing and photos and videos I make and accumulate over the course of my life, deriving meaning from all this stuff, and generally doing things that computers are great at but humans not so much, on my behalf: things I instruct it to do, but also things I haven't yet thought to instruct it to do (but would benefit from it doing).

There are a lot of hurdles to overcome before such a tool could become a practical reality, though, including the financial incentives in this space.

It's no mistake that contemporary voice assistants are riddled with privacy concerns, and my ideal software agent would need to be mine and mine alone for me to be comfortable granting it access to everything I would want it to be able to access, not to mention allowing it to function in my private spaces and moments, serving as an always-on companion.

That said, we're reaching a point where something like this (maybe an Open Source version?) would be technologically feasible, at least at a basic level. And I'm very keen to see what capabilities that might offer us in the coming years.

—

You can support Aspiring Generalist for the cost of a cup of coffee each month ($3/month) via the link below (which also gives you access to periodic bonus emails) or via Buy Me A Coffee.

Or you can support all my work by becoming an Understandary member.