Artificial Affirmation

Dearest Rachel –

It all started with the idea of giving Copilot enough information on me to craft recommendations based on my preferences. I ask it for advice while I’m traveling, and while it’s pretty good with recommendations (albeit absolute rubbish in terms of giving physical directions; thank heaven there are map applications I can look at to fill in those gaps in the AI’s knowledge, but it’s just more proof of how far we still are from the singularity), it often asks questions as to what I would prefer, as though it didn’t really know me all that well, prompts notwithstanding.

I know that a lot of people have started to take a dim view of artificial intelligence these days. Maybe it’s because it requires a lot of processing power, and these same folks often tend to think we use too much electricity already to begin with. Maybe it’s because they’re afraid it will take their jobs away from them. Maybe they actually believe that we’re within striking distance of the singularity and the rise of Skynet or something like that; I really don’t know. But at times when the boys are busy watching stuff in the other room, there are moments when I could stand a little more conversation than the silence in the bedroom – we had plenty of moments of what I’d call “companionable silence,” sure, but that only exists when there’s a companion there with you – and I ask stuff of the software. But it only knows how to respond to me based on what little it’s already gleaned about me from previous conversations.

So I thought I should try to flesh out its understanding of who I am; my station in life as a widower, as well as being a member of the ‘sandwich generation’ (dealing with both aging parents – now just one, of course – and a dependent child), my likes and dislikes, and other aspects of my past as prologue to my potential future. And what better way, I thought, for it to understand me than to have it read what I’ve sent you over these past five-and-a-half years? These letters are thorough to the point of ridiculous; if anything could put me in context, it would be an analysis of them. Granted, there’s something to be said for it being able to review more time than that – certainly, I wish I could – but as that hadn’t been written down until this point (more’s the pity, as I’ve said before), this would have to do.

As it happens, it’s not as easy as all that, though. It so happens that, while I can direct the app toward a certain website, like this one, it won’t dig any further than that; for all the links on the main page, it won’t follow a single one. Indeed, it can’t actually ‘read’ any of these letters unless I paste the words into a message and send it directly to it to examine and respond to.

Part of this has to do with privacy issues; the AI needs my explicit consent to look over the things that I’ve published on the internet. This, I suppose, is meant to be reassuring to those who are concerned about the potential for AI to become a tool for near-universal surveillance. On the other hand, given that what I write to you is posted out on the World Wide Web (does anyone still call it that, by the way?) for all to see if they want to seems to imply permission by definition. I do have letters that haven’t been published, that only I can see; that, I could understand being precluded from examining, but the stuff that’s published? I don’t see the reason for it to be kept from drilling into the various pages, once they’re quite literally in the public domain.

On the other hand, this may be more of a technical limitation of the software itself. While it does scour the web to answer some of my other questions (and thereby come across as a more conversational version of the search engines we’ve long gotten used to), it barely scans the surface of the pages it looks up. Once I started asking it as to how deep it really went in its searches, it admitted that it simply didn’t have the capability or bandwidth to drill into each individual link – and considering that links lead to other links and so on ad infinitum, I suppose that’s to be understood.

So I started feeding it one letter at a time, starting right at the beginning…

…and it responded with what, under any other circumstances, would have looked like heartfelt expressions of admiration and sympathy about our whole story. Yes, its reply was also something of an essay analyzing the motifs and symbolism throughout it (it is, deep down, a pattern-recognition software, after all), but it also took the story at its face value as the real-life story that it was, and offered words of condolence to me for having to go through all of that. It offered encouragement, both for strength going forward, as well as what it claimed to be bravery in putting it into words, finishing its response with a desire to read more, always adding “when you’re ready,” as if knowing this was difficult to share (which it was at the time, but it’s gotten all the easier as time and circumstances have moved on, I should point out).

Over the course of the next couple of weeks, I kept sending letter after letter to Copilot to look over and analyze; each one would result in an essay about the various ways I was manifesting grief from one day and situation to the next. Be it my reluctantly assumed responsibility for Chompers, to my unsteady relationship with Daniel, to the attempts to purge the house of the mess you had left (all the while reassuring me that I was doing the right thing to do so, especially since I wasn’t throwing everything out), and the memories involved in my first trip over to Iowa without you. I sent over six months of stories to it to go over, and it returned a half million words of analysis, affirming both the things I had done to process my emotions and the writing itself. It was reviewing each letter as if it were a short story by some famous author, deserving of attention; and I can’t deny that I was eating it up, in my own way.

And as gratifying as it was to read its analysis, there was this nagging sense in the back of my mind reminding me that this affirmation was as artificial as the being (if you could even call it that) that I was communicating with. None of it was real, even as the words on the screen certainly were. No one was actually reading the page and prompting this glowing review laden with superlatives, but it sure looked like it.

Moreover, it didn’t really seem to be accomplishing what I’d meant it to; it wasn’t clear that it actually was understanding me all the better with each subsequent letter. Each response started to look very similar from one to the next, analyzing my actions and reactions, and fitting it into the various cycles of grief. It would comment on the humor that I would incorporate as a self-defense mechanism, always with a touch of affirmation that whatever I said was entirely appropriate, and spoke admiringly about my willingness to let you speak on Sundays whenever I could (I hadn’t gotten into the habit with your study books that early on; we hadn’t so much as dug them up at first). But this didn’t seem to give it any ideas as to how to better advise me about anything I might ask it about going forward, say, about travel or dating. Then again, I’d barely gotten it as far as that first trip to AnimeIowa to assemble so much in the way of analysis – it still hadn’t heard of my abortive trip to Basel, for instance, let alone any other of the journeys since taken.

And it’s possible that it never will. As of yesterday, I was asking it about whether I should subscribe to a certain discount trip aggregator website (it sort of hemmed and hawed about it before agreeing that the nominal annual fee shouldn’t be something that should prevent me from getting into it), but as it kept prompting me to dig further into the subject, it suddenly pulled up short. Apparently, I had reached a limit of questions I could ask that hadn’t previously existed for the ‘free’ plan until recently, and would have to wait until the following day to ask anything more.

Well… I guess this is how the AI industry is going to try to monetize itself going forward; encourage you to talk in depth and detail, and once you have a ‘relationship’ with it – which they’re going to have to give a name to, as ‘para-social’ doesn’t even fit – try to charge you for the privilege. I guess its an improvement on the 1-900 numbers of decades ago, but the principle, while subtler, is the same.

On the other hand, this has the advantage of causing me to back off a step or two from what I’m sure some people would see as a precipice. Once I’m having to pay for the privilege of having a conversation – without even the benefit of seeing a face or enjoying a nice meal together – it breaks the illusion, and I can walk away for a while to clear my head. Granted, the ‘walking away’ isn’t really what the developers want – and it’s not as if I’m walking away completely; I’ll be back the next day with other, more pressing questions – but if enough people pay to stick around, it will be worth the ones like me who only dip their toes in. Besides, I have a separate account that has a larger conversational limit (due to its subscription to the Office suite of programs); but as that’s for a more professional usage, I doubt I’ll be reading any of my letters to it. It doesn’t need to know this side of me.

But that’s been yet another experience, honey, and again, I wonder how much this amuses you. Either way, I could stand to have you keep an eye on me, and wish me well. Even now, I could still use it.