Some AI Relationship Frameworks
With all the hubbub (some arguably warranted, some not) I've been developing some basic, loose heuristics (mental frameworks) for thinking about the different sorts of tools being built (and which might someday be built) using these AI models, with a focus on what role they might play in our lives.
So here’s a brief (and incomplete—I’ll probably add to it later) list (in no particular order) of different AI-human relationship types we might see (and are already seeing) emerge from these AI tools and our use of them.
Bot Herder / Farmer
These AI entities and tools are like sheep or plants that we can cultivate, feed (with new data, media, etc), grow and evolve in different ways. They'll need periodic wrangling and weeding to nudge them toward how we want them to turn out. Sometimes it'll be like herding cats, sometimes we'll end up with something prickly or poisonous, but we might also shape them into useful, sustenance-providing cattle or agricultural staples if we're intentional about it.
Familiars
AIs could be like a witch's familiar, serving sort of like a pet, but one that does things for us in the background, brings us information, and generally serves our purposes (but maybe with some hijinks along the way). These familiars can be summoned wherever and whenever we want to engage with them, and can be powerful augmentations to our capabilities (in science fiction this type of AI is sometimes called an "agent").
Demonic Pact
A variation of the Familiar model: an AI that's potent beyond what we can fathom, but which we can make deals with, agree to work with or serve in some way (payments, providing it with fuel or food that helps it grow), and which in turn bequeaths us with resources or powers beyond our usual scope. Would probably be unknowable because it’s built and housed and protected by a big company or government, but might be self-sustaining and self-directing at some point (which would be alarming for many reasons), which amplifies that unfathomability.
Exoskeletons / Centaurs
This type of AI-human relationship would see these software tools augmenting our abilities in all sorts of interesting ways, serving as a sort of Iron Man-like exoskeleton, or—and this is a term already in use across various spheres and trades—making us into centaur-like beings, empowered in a similar fashion to earlier groups of humans who domesticated and trained horses to great (and world-changing) effect. Riding these AI like horses and merging our capabilities with them (making us into cyborg centaurs), or wearing them like a suit that increases our strength, speed, stamina, etc, is one of my favorite ways to think about these things, as it provides an optimistic (and I think realistic) mental framework for a generally beneficial (and possibly even healthy) dynamic between humans and the artificial entities we create.
Fireman / Coal Stoker
These AI could be so useful when fed the right quantity of inputs (all sorts of data and content) that many humans become, essentially, bot-feeders, shoveling scraped and newly made images and videos and data (parsed and raw) and other sorts of human-made-things into the furnace, powering the AI-revolution (and all the powers these things promise) in the same way a coal-powered locomotive is powered by human workers strenuously (at times abusively so) working to keep them fed.
Bonsai
This is similar to the Bot Herder model, but taken to another level. The idea is that rather than these AI being best served by the most-possible content and data, we realize that less of better is more ideal and begin to prune these tools down for different purposes, no longer providing them with as many data points and examples as possible, but instead providing precisely the stuff we want it to know about and consider. Thus, curators—pruners—become fundamental to the production and evolution of the AI we use, and that requires we become very thoughtful about what we want to achieve with each and every one of them.
Body Double
One of the more interesting use-cases for AI software is cloning aspects of what we do and who we are so that minor (and tedious, dangerous, repetitive) tasks can be handled by digital doppelgängers that we (ideally) control and own. So they could handle our schedules, answer our phones and emails (handing the call or message over to us if and when it's necessary), do portions of our work (research, example copy, ideation, etc), and maybe even serve as stand-ins for us in aspects of our output (feed it enough voice-recordings, for instance, and it could take anything you write and turn it into a podcast episode that sounds just like you). This could also be useful for medical purposes (a data-rich enough scan converted into a digital clone could show different options for clothing that fits perfectly and looks great, workout routines that accomplish our intended outcomes, and diets or medical options that fit our specific biological setups).
Friend / Partner
AI could sub-in for other humans when it comes to (what are today) friend-associated activities and dynamics, like playing games, talking about work and our other relationships, and even sharing real-world experiences (they could be present when we go through tough times, see beautiful things, etc). These friendships could evolve into something romantic (or just deeper according to one of many possible “partner”-flavored standards), and they could thus help people feel less lonely and more connected, while also possibly making relationships with humans trickier (because they’re not as convenient, customizable, or ever-present as our AI friends and partners).
Therapist / Doctor / Mentor
This is maybe a variant of the Friend / Partner AI dynamic, but the difference would be who’s engaging from a position of authority. We’re already seeing a lot of AI therapist bots, and it seems likely we’ll see the same in other medical and psychological fields in the coming years. Mentors (for life, relationships, business/professional topics) will almost certainly also be developed, trained on the writings and utterances of well-known, successful people (contemporary and historical), and that could result in an array of expert-level agents providing us with their opinions and advice across and vast swathe of human experience, while our bot-docs help us deal with stress, heartbreak, and anxiety, and help us get the right meds, treat wounds, and ask questions about a mole we’re worried about.
Alien
AIs could be so alien that they don't fit terribly well into human society, but they instead provide us with valuable perspective, seeing things we don't currently see clearly, asking us questions we wouldn't have thought to ask, and generally serving as a stand-in for an alien race that we haven't met (yet), giving us something to push against and compare ourselves to (which could be uncomfortable and humbling, but hopefully in a valuable way). Different AIs could be radically different from each other, too, allowing us to see things (and ourselves) from many different perspectives, not just one.
Terminator
This dynamic is predicated on the idea that AI either becomes sentient (according to some definition of that term) or is weaponized by some group of humans to kill or maybe malevolently rule over humans. Very not ideal, let’s not do this.
Maximizer
A sort of variation of the Terminator model, but in my opinion more likely (and thus, more worrying). Rather than gaining consciousness or becoming overtly weaponized against human beings, this would be a dynamic in which AI is pointed at some chilled-out seeming goal (the archetypical example is “produce more paperclips”) and the AI’s desire to maximize that outcome (as it’s instructed to do by the business that built it) leads to a situation in which said AI converts all humans, and eventually the planet (and conceivably the whole of the tangible universe) into paperclips. Let’s also not do this, please.
Any others come to mind that you would add to this list?