Artificial "Intelligence"
From business to the cultural zeitgeist, there’s a lot of buzz around about AI. Even a push for it in business (that goddamn “I just Mailchimp’d my marketing!” ad that sounds like a great way for shitty companies to further blow up my inbox with unwanted spam comes to mind) and fears of how it could put people out of work.
And I can’t help but think, “Are the people considering that fuckin’ high or something?”
Or, alternatively, “Where do they get their weed?”
AI’s annoying, and I’m a little burned out on repricing books and whittling my way through The Decline and Fall of the Roman Empire (written around the Founding Era, quite intriguing to reflect during a Falling Era) so I’m here to gripe about it.
I guess I’m a few weeks late on Edolf getting Grok to declare itself “Mecha Hitler.” I see he claims Grok was “manipulated” into doing this, which is shockingly honest for him, considering he’s the guy behind the scenes demanding the changes to the code that made the bot become that. It’s like we’re living through a really shitty Terminator origin story, if you’ve ever wondered why an AI like Skynet would decide to genocide the fucking planet. Apparently the only change Edolf forgot to insist on was, “Don’t say the quiet part out loud.”
AI also is the reason I’ve been getting half a dozen or more calls, daily, about how the most recent republican bill has either expanded or limited my (nonexistent) Medicare coverage. These automated voices can’t decide, but then, that’s how scams operate and I imagine a lot of senior citizens are getting calls like this, too.
I’m actually confident of it because local news has been giving warnings about these AI scams that are targeted to take advantage of vulnerable seniors.
And more companies are adopting human-like voices to operate menus. You used to be able to bypass this by using words like human, operator, live person, representative, agent, or pressing 0.
Must’ve made getting to customer service too easy, now some of these places will insist you provide a “brief explanation” or “in a few words” tell them what you need. But then they hear the bathroom fan or TV or Charlie walking instead of my voice and everything goes haywire. Or it is confused because, again, AI is not advanced enough to be rolled out this way; it’s very buggy. My insurance company’s menu works this way, so did the plumbers I had to call out on the fourth; 3/3 on buying a new place and needing plumbers within a month; thanks, I hate this cosmic joke.
You think I want to spend five minutes dealing with some shit automated menu only to find out you’re not even open I just didn’t realize it was a bot? You take me on this roller coaster ride of scheduling a plumber and then tell me you’re not open? My shitter just backed up when I took a shower, I don’t need a plumber Monday, I need one today.
What was so wrong with the original menu where Business Hours was “Press 2” or “We Are Not Open Today”?
A perusal through free audiobooks online (praise be to LibriVox) shows that a lot of people use AI voices to create ‘new’ ones, slap ‘em on YouTube, and then try to get ad revenue. Not a bad idea, I can’t condemn, I’d probably have done the same if I thought of it first, but it looks like YT has caught on to some of them and nipped out ads so people can’t do that as easily.
While I find this a more acceptable and beneficial use of artificial intelligence, it also does highlight some of the weaknesses that the technology struggles with and which clearly outs it as inferior to human capacity.
A few things I’ve noticed: AI has a difficult time pronouncing numbers properly or in the same way a person would. Even attempts to emulate a more colloquial style (twenty twelve for 2012) have a weird, awkward extended pause between twenty and twelve, unlike a person. But it also sounds odd when you hear AI pronounce it as, “In two-thousand and twelve…” because nobody talks like that, at least, no one who speaks English natively I’ve heard.
(AI struggles with counting in general, as Microsoft Rewards via Bing reminds me daily. Their mobile app should require twenty searches to get to 100 points per day. I check my searches against what is counted. 3/11, 3/8, 1/4, 2/5, and other such ratios are the number of “counted” vs. “made” searches for some of the last few days. Who hasn’t had a boss that ignores a chunk of your work and demands you do more? Well, now we’re programming those assholes into AI so that we can be further taken advantage of by shitty tech bosses. At my scam job, the counter that clocked length of time on the phone and calls made worked similarly. Usually, I’d be at 450-500 calls before it’d register the daily requirement of 200).*
Particular words stand out, as well. Whether you are talking about a group as it “wound its way down a hill,” or someone cutting their hand while chopping veggies and getting a “wound,” it will be pronounced in the context of the injury pronunciation. It is jarring to the ears. (I also have come across one AI voice that did the opposite and everything “wound its way” pronunciation, even deep cuts or wounds).
In AI world, too, you do not “resume” a task or a journey; you “résumé” a task or a journey. If you’re looking for a “minute difference,” you probably mean an extremely tiny difference, not the interval of sixty seconds which AI will pronounce.
The AI also outs itself in audiobooks, and the creators out themselves as not having done their editing work, because of what I call ‘floating page numbers.’ If you check websites with public domain texts, there will occasionally be numbers left in the text which signify the page numbers of the original edition that was used to transcribe the work.If you listen to an AI audiobook you 12just might find that there are random13 number scattered throughout the 14pages and so out of nowhere15 you will suddenly hear a number pronounced without context or relevance.
When it comes to non-audiobook AI, I’ve noticed a tendency of repetition in what appear to be AI-generated essays based on a prompt. I think I get a glimpse at what teachers mean when they talk about AI essays being noticeable. They follow a sort of high school hamburger essay approach, particularly noticeable in the repetition I think, but the vocabulary doesn’t quite line up. It’s not quite what you’d expect from a smart teen or from a simplistic adult, it’s just… weird and kinda off-putting after a while.
(The high school essay also applies in that AI, similar to what my sophomore English teacher taught us about WASL graders, is incapable of differentiating fact from fiction and weighing the validity of sources. Kinda like my complaint on Warriors of God: AI will slip in anecdotes or folklore around a topic as if it is fact rather than speculation. Unlike the author of the book mentioned, however, it is not done for a fun glimpse into how people thought; it is presented as studied fact, which it is not).
I could be completely off on that, but anyway, there’re people out there who can easily point out examples of internet comments or AI-generated images/attempted Deep Fakes (with six fingers, often, if you check), both of which are things I still struggle to catch myself.
It’s a brave new world—and with such people and not-people in it, I don’t know what to make at times.
—
*I did pick out some of my favorite numbers to call. The weird thing is sometimes a long pitch with a customer wouldn’t count, but “The number has been disconnected or is out of service,” would count. My favorite, though, to rack up call length was a guy who had a “You just won a $100 Wal Mart Gift Card!” scam. I figured we both had similar requirements for time/calls, so I called this guy back a few times and we’d throw our pitches back and forth. Then he got irate and hung up on me on like day four. Asshole.
—Speaking of people, I’ve ducked two insurance scams from real people, not AI not-people, recently. Thank god I didn’t have cash on me then or I’d be out it, cause No Contact Claim One was 0% on me. The person was still brazen enough to text me after insurance’s decision and ask if I’d pay out of pocket for the rim of their luxury car cause they drive like an idiot, which is Frank Gallagher level shameless.
Claim Two is—well, it’s not filed and if he had a valid claim I’d knocked his bumper loose, it’d be filed and there’d be at least a scuff in my paint (we had parked very close), he wouldn’t have waited over three weeks to ambush me in the parking lot demanding cash. He can fuck right off with his messages about making it a “bigger deal” how he’s “gonna be around” and “wants to play.” Cops didn’t want to show up and take the report, so I guess that means I’ll just beat the shit out of him with a tire thumper if he tries to come at me aggressively with threats again. (Or, more my style, keep my car in the garage that came with my purchase—I do hate backing in, but I haven’t hit any walls yet; good thing that dumb fuckhead doesn’t know my garage is manual-open no-lock and I don’t have the cash to fix it).