Stared into the Sun
I spent the last couple of days in (or more accurately, sort of near but not really) Zürich, where my company's main headquarters is located. This was for an SEO workshop in theory, but in practice it was a bunch of people sitting around talking about the customer journey which, without going into too much detail about what my employer does, is extremely boring shit. For reasons based soley around concepts such as "not wanting to be broke and homeless" and "wanting to be able to continue living in Germany, which requires me to have a job," I summoned the required social energy to engage in these discussions and take them seriously and bring the eleven (fuck that is a long time) years of job experience to bear on the discussion. In spite of everything, I do somehow manage to pull off something like professionalism now and again.
This does not, unfortunately, include taking the topic of using our in-house fork of ChatGPT in order to improve my efficiency or whatever. I don't like AI for a whole bunch of reasons that range from the moral (they're built on the theft of work, they're being used as excuses to lay off giant swaths of people, they are being used to flood the internet with slop that has accelerated the enshittification of everything to lightspeed, they require gigantic amounts of energy and water which makes them like an environmental disaster in a world that is in the middle of a slow-roll climate apocalypse) to the more practical (they do not work that well) to the more personal (I am a professional writer and I will cut my own fucking throat before I turn to an LLM to write something for me). Whenever the topic of using AI comes up, I bite my tongue and more or less just refuse to use the thing, in spite of the constant suggestion from my boss that it's a fine tool that will make my life easier.
Unfortunately, while it is easy to not use our AI chatbot (or whatever the fuck you want to call these things) in my day-to-day, it is significantly more difficult to not use it when you are in a workshop about using it. So in spite of my best efforts, I Used a Damn AI Bot for the first time.
There is a version of this post that exists in a different universe, maybe, where this is the part where I talk about how the technology won me over. "It turns out it really was easy to use," this alternate-universe me who sucks would say. "I asked it to make a list of products and then write an email making a recommendation as to the best one of the products in said list and it did that," I might add, "I was very impressed."
Instead I have to report the simple fact which is the whole think felt like the writing equivalent of drinking a Health Tonic bought off the back of a wagon from a man in a tattered top hat and not being killed by it or anything but also not having any sort of positive health experience as a result. I did ask it to do more or less what I described, and it did generate me a list of stuff and then make a recommendation in email format (because I suffer from the inability to take this shit seriously, I also instructed it to write the email in the style of e.e. cummings - which if you think that the style of e.e. cummings means "no punctuation or capitalization and also it's a poem" and nothing else then sure, I guess it did that too), but when we repeated the question it gave us a different answer, and after about four or five additional prompts the thing had convinced itself that I was named Jeff and also it forgot to put our company's product in its list of rankings again.
None of it was what I could call accurate - it seemed legit, if you didn't interrogate it any further, but that was it. As soon as you start asking the questions of "where did this information its giving me come from" you find out the answer is more or less "we just pulled some shit at random." Hearing the presentor of the workshop talk about how he'd used AI to design his company logo and also regularly used it to generate short promotional videos for clients made me write a note down that read simply "if someone I was working with did that shit I would fire them immediately." The video he showed off looked bad and in spite of being like 15 seconds long managed to still break down by the end - the subject at the center of the video just started vibrating for no reason, and the background got smeared together. It sucked, but we were being told this was The Future of Content Creation anyway.
"Be sure to be specific in your prompts," we were told, "or the AI will hallucinate," which is the word they use instead of saying the truth which is the AI will fail to work the way they say it works. I guess "hallucinate" sounds better than "fuck up," even if that's more accurate. The only thing that bothered me more than this overly-cutesy way of describing "the technology did not do what it is supposed to do" was the way that as soon as you poked at the results even a little it became clear there was no way to actually trust what you were being given without spending a bunch of time doing the very work that AI was supposed to do in order to check the work.
Like we're a marketing team being told "you can use AI to do your market research" but also being told "be sure to check everything the AI gives you to be sure it didn't hallucinate" except that
- AI does not give you sources for where it pulls its stuff unless you specifically tell it to source its material, and even then it might not do that for everything.
- The technology as presented to us in the workshop was "think of how easy and time-efficient this is" and brushed off any hallucinations as being the fault of the user and also they didn't happen that often right of course not.
- If the workflow for "responsible" use of AI is "ask it a question and ask for sources, then manually go check those sources to make sure they're accurate and legit" then we've really just reinvented using search engines so what the fuck are we even doing?
I guarantee you most people aren't going to do that legwork, because the part that tells you to do that legwork is in tiny text at the bottom of the chat window and we're all too fucking busy to do that anyway.
So a waste of time, really, even from a research standpoint. I cannot get the appeal - a coworker said his daughter uses ChatGPT instead of google now and all I could think was that it is a terrible way to use the internet. Not that most ways of using the internet aren't terrible, but you know what I mean.