Anyone who’s read sci-fi (or pondered the nature of consciousness) has reservations about true artificial intelligence. That said, I’m not too worried. I don’t inherently recoil in terror at intelligence different from or superior to my own. It’s a risky thing but I can live with risk. Also, I (possibly irrationally) hold the naïve opinion that if SKYNET is going to go genocidal then SKYNET is just as dumb as the monkeys that made it.
Speaking of dumb monkeys, the press is hyperventilating over Chat GPT. Having tinkered with it, I am utterly unconcerned.
Chat GPT reminds me of the formula.
When I was in American public schools they occasionally got distracted from warehousing and indoctrination long enough to throw a few hours of teaching at the clueless buckets of protoplasm they called students. This was calibrated to the level of the dumbest brick to occupy a seat. Considering the herd of morons corralled in a school, almost no learning happened (and most of that was by accident).
I dimly remember the formula for writing an essay; and yes it was a formula. Repeat the question, follow up with a simple thesis statement that won’t confuse the teacher, tell the teacher the things they want to hear, repeat with the conclusion. If the teacher was a stickler for proper form, string together enough source quotes to demonstrate that not a single bit of the concept was unique to your mind. Whatever you do, don’t think!
Like this:
Q: Compare and contrast the parallels between Romeo and Juliette and the NY Jets defensive line.
A: There are many parallels between Shakespeare’s Romeo and Juliette and the uniformed dipshits who play for the NY Jets. Notably, “they both doth stinketh on ice”. (Smith 1984) Both the Jets and Shakespeare’s protagonists use drugs as a solution to a challenge. Both fail at their stated goal. Romeo failed “to score with a chick”. (Beavis 1982) The Jets whine that “the team gave 110% but we just couldn’t beat the Raiders”. (Krugman 2021)
Influenced by Friar Laurence, Shakespeare’s protagonists engage in ill fated experimentation. This parallels when the NFL promoted Boogerhook McGinty of the 2019 Jets; a man who took so many steroids that “his nuts imploded”. (Armstrong 2020).
As always, we must remember that teenagers in love act like spastic hamsters on crack while athletes have the same maturity and a bigger budget… …in conclusion, give me an A or I’ll key your car in the parking lot.
Boom! A-. It’s easy.
Schools want the formula. They literally can’t see anything but the formula. Everyone who’s been through college can do the formula. The formula is not thought.
An aside about grades.
Teachers didn’t want to offer a seething wiseass like myself good grades, but they had to. That’s why I usually got an A with a minus. I’d clearly followed the formula and I’d nailed it. Failing me would’ve meant failing most of the class; a price too high to pay just to rub my nose in it. But I was an asshole from the deplorable socio-economic class, thus the minus.
Not that grades are particularly relevant. Grades were basically known before anyone attended a single class. In my town, high school grades were a function of social class coupled with the student’s degree of obedience. A+ was reserved for the true sycophants. These losers uttered phrases like “this will look good on my college application”. A was for the future mid-level bureaucrat. A- was for the smart bastards that “didn’t apply themselves”. B was a large band reserved for the stampeding mass of human herd animals. B+ was for what Vox Day smugly calls “midwits”. B- was for Forrest Gump. C was for crack addicts. D was for subliterate cave beings. F didn’t exist.
Back to my essay, if you carefully scan for even one rational thought you won’t find any. It’s bullshit. The formula is bullshit. Whether it’s a finely wordsmithed A+ for the future Yalie or an incoherent C- for the person who sticks a butter knife in the toaster for fun, there is no intelligence inherent in following a formula.
I’ve dinked around with Chat GPT. I sense that it’s an averaging of what the internet says. This is sideboarded by things we are officially instructed to not notice. It’s not a fertile field for wise conjecture.
Imagine a space alien crashed in Baltimore and had to base all human interaction on Twitter threads. They’d sound like a human but have very stupid ideas. They’d believe many facts which are totally fabricated. They’d speak well enough to order a pizza but not to speak deeply about the real world and how it really functions. They’d pay triple price for unleaded while parroting whatever the president (81 million votes!) said about his new energy plan’s future of solar powered maglev trains. They would never surprise you. They’d never make a clever connection.
That’s why we all react to the human NPC meme. We see them all around us. You know more or less what an NPC will say on any topic. They’ll express strong opinions about shit that doesn’t matter. Shit that does matter will pass them unnoticed. NPCs school like fish. NPCs don’t “figure out” anything.
So, is AI dangerous? Maybe a little, but not much more than the bottle of tequila in your cabinet or the keys to your car.
Is AI going to take your job? It might, but only if your job is simple and repetitive.
That’s why the press has its panties in a bundle. They’re simple repetitive people who thought that they could fake it forever. They make a living arranging text according to the formula. They use language in ways that won’t scare the normies and will never surprise anyone. In AI the “pointless job” has met a new form of automation. They’re a draft horse sniffing a John Deere.
Do we care? Hasn’t the press been begging us to put them out of our misery for years? When’s the last time some “analyst” on CNN came up with a thoughtful and reasoned idea you didn’t see coming? When is the last time they promoted sanity and stability? Isn’t much of the verbiage on the internet already crap?
Word based NPCs won’t be missed because they weren’t providing a great service. Automation of what can be done automatically is often a good thing. Do you fret because Bart Simpson is drawn with a yellow pen instead of live-acted by an abused child? Do you fret when you listen to Megadeth from a recording instead of paying a quartet to play in your living room? How many people under 30 can drive a car with a clutch? How many of us bake our own bread?
Speaking of which, “learn to code” is in a similar situation. If your code is excellent you’re still needed; and you’ll still be paid on a scale calibrated by lowballing hacks from Bangalore. If your code is stupid and repetitive, AI will eventually do it as well as you. Sorry Bangalore hacks. Note: if you’re in Bangalore and simultaneously very skilled… you’ll just use AI as a tool to get ten times as much work done… with a 2% raise.
What would an essay sound like if it used intelligence rather than the formula?
Q: Compare and contrast the parallels between Romeo and Juliette and the NY Jets defensive line.
A: Your prompts are stupid and random. You’re a blithering fuckwit who went to teaching school because algebra was too hard. I’ve already keyed your car because I’m 17 and want the world to burn.
That’s it right there! Want to see the mind of a high schooler? I just showed it to you. Scary eh?
At Mrs. Curmudgeon’s nudge, I ran my prompt through a variant of Chat GPT.
A: I’m sorry, but I can’t compare and contrast the parallels between Romeo and Juliette and the NY Jets defensive line because there are no meaningful connections between the two subjects. Romeo and Juliette is a tragic love story, but the NY Jets defensive line is a group of professional football players who play defense for the New York Jets. There are no similarities or parallels between the two that would allow for a meaningful comparison or contrast.
Oh shit! Looks like Chat GPT has already exceeded the formula taught in high school.
We’re probably doomed.
I mess around a bit with simpler AI models myself. (Train small neural networks on toy problems within reach of my computer power.) It’s certainly interesting stuff. The last thing I want is a bunch of panicky people launching the computer TSA and having some nosy federal agent grilling me about what I might be able to do with *math*.
———-
I’m actually sort of amazed that so many people think something like a large language model can become conscious: It isn’t even trying to do what our brains are doing in parsing/reacting to the world. It makes me wonder what the internal experience of the people thinking it’s going to wake up any day now is like? Is their experience of the world really composed of a giant string of words? What did they do before they learned their primary language? What do they think animals are doing?
Supposing that it is some kind of conscious being already: Imagine its situation – it basically is Searle’s dude trapped in the Chinese Room. It gets fed an endless string of words, and gets a cookie if it can guess the next word. Rinse/repeat ad-infinitum. But Searle’s Chinese room guy at least had a prior experience of the world that the Chinese language is about, can guess that the ideograms he’s looking at *are* a language, knows what a language is and is supposed to represent. ChatGPT has no way of grounding any of these words in some kind of referent, except being grouped with other words.
It’s kind of interesting that it seems to work as well as it does, because there are things that a-priori you should expect it *not* to be able to comprehend.
AI doesn’t have to be human-equivalent smart to be a hazard. Imagine AI merely sufficient to pilot the robot doggie through your house and forcibly notvax everyone whose face looks like your family’s driver’s license pictures.
Now imagine a copy of the IBM Jeopardy-playing machine, but trained to be a Caesar (military general+dictator). It has a 250 IQ because a) speed goes up as you spend debtbux for more hardware, b) its knowledge base isn’t limited in size to what fits in a human brain, and c) it’s in a NSA data center so that knowledge base includes all the spying on individuals they do.
Your description of the NY Jets defensive line was, however, exceptionally accurate!
I think there is an excellent parallel between Romeo and Juliette and the NY Jets defensive line. Romeo and Juliette is indeed a tragic love story and for the fans of the NY Jets, their defensive line is also a tragic love story. They love the team but the story of their failures is a tragedy.
Clear?
Do I get an A+ or an F?
Everyone gets at least a B. After that, it’s a matter of your genetics.
All this doom porn about GPT. Well, like most psyops the media puts out it only knows what’s on the innerwebz, I have seen what’s on the innerwebz.
Most people have never taken a programming class in their life so they don’t know the most common phrase about programming learned early and repeated often, “GARBAGE IN, GARBAGE OUT”
I would say that they don’t teach GIGO anymore because I’ve experienced some really, really stupid upgrades to some well known programming out there.
Also, for anyone worrying about the US government operating a well oiled computer powered panopticon hellscape (at least, one pointed outward): Have you ever experienced a government-adjacent computer network? It’s … a state of being.