Blog Archives
Dead Internet
There are two ways to destroy something: make it unusable, or reduce its utility to zero. The latter may be happening with the internet.
Let’s back up. I was browsing a Reddit Ask Me Anything (AMA) thread by a researcher who worked on creating “AI invisibility cloak” sweaters. The goal was to design “adversarial patterns” that essentially tricked AI-based cameras from no longer recognizing that a person was, in fact, a person. During the AMA though, they were asked what they thought about language-model AI like GPT-3. The reply was:
I have a few major concerns about large language models.
– Language models could be used to flood the web with social media content to promote fake news. For example, they could be used to generate millions of unique twitter or reddit responses from sockpuppet accounts to promote a conspiracy theory or manipulate an election. In this respect, I think language models are far more dangerous than image-based deep fakes.
This struck me as interesting, as I would have assumed deep-faked celebrity endorsements – or even straight-up criminal framing – would have been a bigger issue for society. But… I think they are right.
There is a conspiracy theory floating around for a number of years called “The Dead Internet Theory.” This Atlantic article explains in more detail, but the premise is that the internet “died” in 2016-2017 and almost all content since then has been generated by AI and propagated by bots. That is clearly absurd… mostly. First, I feel like articles written by AI today are pretty recognizable as being “off,” let alone what the quality would have been five years ago.
Second, in a moment of supreme irony, we’re already pretty inundated with vacuous articles written by human beings trying to trick algorithms, to the detriment of human readers. It’s called “Search Engine Optimization” and it’s everywhere. Ever wonder why cooking recipes on the internet have paragraphs of banal family history before giving you the steps? SEO. Are you annoyed when a piece of video game news that could have been summed up with two sentences takes three paragraphs to get to the point? SEO. Things have gotten so bad though that you pretty much have to engage in SEO defensively these days, lest you get buried on Page 27 of the search results.
And all of this is (presumably) before AI has gotten involved belting out 10,000 articles a second.
A lot has already been said about polarization in US politics and misinformation in general, but I do feel like the dilution of utility of the internet has played a part in that. People have their own confirmation biases, yes, but it also true that when there is so much nonsense everywhere, that you retreat to the familiar. Can you trust this news outlet? Can you trust this expert citing that study? After a while, it simply becomes too much to research and you end up choosing 1-2 sources that you thereafter defer to. Bam. Polarization. Well, that and certain topics – such as whether you should force a 10-year old girl to give birth – afford no ready compromises.
In any case, I do see there being a potential nightmare scenario of a Cyberpunk-esque warring AI duel between ones engaging in auto-SEO and others desperately trying to filter out the millions of posts/articles/tweets crafted to capture the attention of whatever human observers are left braving the madness between the pockets of “trusted” information. I would like to imagine saner heads would prevail before unleashing such AI, but… well… *gestures at everything in general.*
The Nature of Art
The following picture recently won 1st place at the Colorado State Fair:
Don’t know about you, but that looks extremely cool. I could totally see picking up a print of that on canvas and hanging it on my wall, if I were still in charge of decorating my house. Reminds me a bit of the splash screens for Guild Wars 2, which I have always enjoyed.
By the way, that picture was actually generated by an AI called Midjourney.
Obviously people are pissed. Part of that is based on the seeming subterfuge of someone submitting AI-generated artwork as their own. Part is based on the broader existential question that arises from computers beating humans at creative tasks (on top of Chess). Another part is probably because the dude who submitted the work sounds like a huge douchebag:
“How interesting is it to see how all these people on Twitter who are against AI generated art are the first ones to throw the human under the bus by discrediting the human element! Does this seem hypocritical to you guys?” […]
“I’m not stopping now” […] “This win has only emboldened my mission.”
It is true that there will probably just be an “AI-generated” category in the future and that will be that.
What fascinates me about the Reddit thread though, is how a lot of the comments are saying that the picture is “obviously” AI-generated, that it looks shitty, that it lacks meaning. For example:
It reminds me of an article I read about counterfeit art years ago. Most of the value of a piece of artwork is tied up into its history and continuity – a Monet is valuable because it came from Monet’s hand across the ages to your home. Which is understandable from a monetary perspective. But if you just like a Monet piece because of the way it makes you feel when looking at it, the authenticity does not matter. After all, most of us have probably only seen reproductions or JPEGs of his works anyway.
At a certain point though, I have to ask the deeper question… what is a “Monet” exactly?
Monet is rather famous, of course, and his style is distinctive. But aside from a few questions on my high school Art exam decades ago, I do not know anything about his life, his struggles, his aspirations. Did he die in poverty? Did he retire early in wealth? Obviously I can Google this shit at any time, but my point is this: I like The Water Lily Pond. The way it looks, the softness of the scene, the way it sort of pulls you into a season of growth you can practically smell. Who painted it and why couldn’t matter less to me, other than possibly wanting to know where I could find similar works of this quality.
This may just say more about me than it does art in general.
I have long held the position that I do not have favorite bands, I have favorite songs. I have favorite games, not studios or directors. I have favorite movies, not actors. Some of that is probably a defense mechanism – there are many an artist who turn out to be raging assholes, game companies that “betray” your “trust,” and so on. If part of the appeal of a given work is wrapped up in the creator(s), then a fall from grace and the resultant dissonance is a doubled injury. Kevin Spacey is not going to ruin my memories of American Beauty or The Usual Suspects, for example. I may have a jaundiced eye towards anything new, or perhaps towards House of Cards if I ever got around to watching that, as some things cannot be unlearned or fully compartmentalized (or should be).
So in a way, I for one welcome our new AI-art overlords.
Unlike the esteemed Snoo-4878, I do not presume that any given human artist actually adds emotion or intention into their art, or whether its presence enhances the experience at all. How would you even know they were “adding emotion?” I once won a poetry contest back in high school with something I whipped up in 30 minutes, submitted solely for extra credit in English class. Seriously, my main goal was that the first letter of each line spelled out “Humans, who are we?” Granted, I am an exceptionally gifted writer. Humble, too. But from that experience I kind of learned that the things that should matter… don’t. Second place was this brilliant emo chick who basically wrote poetry full-time. Her submission was clearly full of intention and personal emotion and it basically didn’t matter. Why would it? Art is largely about what the audience feels. And if those small-town librarians felt more emotions when hit by big words I chose because they sounded cool, that’s what matters.
Also, it’s low-key possible the emo chick annoyed the librarians on a daily basis, Vogon-style, and so they picked the first thing out of the pile that could conceivably have “won” instead of hers.
In any case, there are limits and reductionist absurdities to my pragmatism. I do not believe Candy Crush Saga is a better game than Xenogears, just because the former made billions of dollars and the latter did not. And if the value of something is solely based on how it makes you feel, then art should probably just be replaced by wires in our head (in the future) or microdoses of fentanyl (right now).
But I am also not going to pretend that typing “hubris of man monolith stars” and getting this:
…isn’t impressive as fuck. Not quite Monet, but it’s both disturbing and inspiring, simultaneously.
Which was precisely what I was going for when I made it.
Zombie Smarts
I have been playing some more 7 Days to Die (7DTD) now that the Alpha 17.1 patch came around. There have been a lot of tweaks to the progression mechanics and Perk system, including some level-gating on Iron/Steel tool recipes. The biggest change, however, was to zombie AI.
In short, zombies are now impossibly smart… and impossibly dumb.
It’s been long enough that I don’t even remember how zombies behaved in prior patches. What zombies do now though, is behave in perfect tower defense intelligence: the shortest distance between them and you, with walls adding a virtual number of steps. Zombie are also perfectly prescient, knowing exactly which wall blocks have the lowest remaining health, and will attack that spot en mass to get to you. At the same time, zombies prefer not attacking walls to X extent, if they can walk there instead.
The result? Cue the Benny Hill theme:
Essentially, the current 7DTD meta is to not create bases at all, but rather mazes that funnel zombies into kill zones and/or large drops that loop them around until they die of fall damage. The devs have added a “zombie tantrum” mechanic to try and get some damage on looped mazes – zombies will attack anything nearby when they fall, possibly weakening your support pillars – but that will be metagamed away with multiple platforms or deeper holes.
To be clear, the prior zombie meta was solved by simply building an underground bunker. At that time, zombies could not dig into the ground, and disregarded the Z axis entirely – it was possible to hang out in the middle of a bridge and often have a nice grouped pile of zombies directly below you to hit with a Molotov. I played the game enough to recognize which Point of Interest had a pre-built bunker located underneath it, and often sought it out immediately after spawning so I could all but ignore the titular 7th day horde attack.
That said, how smart should zombies be?
The only way to answer that question is to ask what the game you’re making is supposed to be about. When you add tower defense mechanics, you get a tower defense game. This will preclude people from building nice little houses in the woods, and instead opt for mazes and obstacles and drops. It becomes a much more technical game, solvable with very specific configurations. Having dumber zombies frees up a lot more base designs, on top of possibly requiring a lot more attention to one’s base after an attack, as a single “dumb” zombie could be weakening a support in an unused corner.
My initial “solution” would be to mix and match, but I think that’s actually the worst of all possible worlds. Instead, I think zombies work best as environmental hazards. Bunkers might make you invulnerable to nightly attacks… but you have to leave sometime. Shouldn’t the punishment for hiding underground be the simple lack of information of what’s going on, combined with having to spend your morning hours slaying the zombie hordes milling about outside?
I guess we’ll see what the devs eventually decide. At present, there simply seems to be a maze-based arms race at the expense of any sort of satisfying nesting. If the 7DTD devs want to double-down, well… thank god for mods.
Dynamic vs Random
Keen has another post up lamenting the stagnant nature of modern MMO game design, while suggesting devs should instead be using ideas from games that came out 15+ years ago and nobody plays today. Uh… huh. This time the topic is mob AI and how things would be so much better if mobs behaved randomly dynamically!
Another idea for improving mob AI was more along the lines of unpredictable elements influencing monster behavior. “A long list of random hidden stats would affect how mobs interact. Using the orc example again, one lone orc that spots three players may attack if his strength and bravery stats are high while intelligence is low. A different orc may gather friends.” I love the idea of having visible cues for these traits such as bigger orcs probably having more bravery, and scrawny orcs having more magical abilities or intelligence — intelligence would likely mean getting friends before charging in alone.
The big problem with dynamic behavior in games is that it’s often indistinguishable from random behavior from the player’s perspective. One of the examples from Keen’s post is about having orcs with “hidden stats” like Bravery vs Intelligence that govern whether they fight against multiple players or call for backup. Why bother? Unless players have a Scan spell or something, there is no difference between carefully-structured AI behavior and rolling a d20 to determine whether an orc runs away. Nevermind how the triggers being visible (via Scan or visual cues) undermine all sense of dynamism. Big orc? Probably not running away. If the orc does run away, that’s just bad RNG.
There is no way past this paradox. If you know how they are going to react based on programming logic, the behavior is not unpredictable. If the behavior is unpredictable, even if it’s governed by hidden logic, it is indistinguishable from pure randomness. Besides, the two absolute worst mob behaviors in any game are A) when mobs run away at low health to chain into other mobs, and B) when there is no sense to their actions. Both of which are exactly what is being advocated for here.
I consider the topic of AI in games generally to be one of those subtle designer/player traps. It is trivially easy to create an opponent that a human player could never win against. Creating an opponent that taxes a player to their limit (and not beyond) is much more difficult, and the extent to which a player can be taxed varies by the player. From a defeated player’s perspective, there is no difference between an enemy they aren’t skilled enough to beat and an unbeatable enemy.
You have to ask yourself what you, as a hypothetical designer, are actually trying to accomplish. That answer should be “to have my intended audience have fun.” Unpredictable and tough mobs can be fun for someone somewhere, sure, but as Wildstar is demonstrating, perhaps that doesn’t actually include all that many people. Having to memorize 10+ minute raid dances is bad enough without tacking convoluted mob behavior outside of raids on top. Sometimes you just want to kill shit via a fun combat system.
Themepark MMO players enjoy simple, repetitive tasks – news at 11.
In An Age 
ChatGPT
Jan 26
Posted by Azuriel
Came across a Reddit post entitled “Professor catches student cheating with ChatGPT: ‘I feel abject terror’”. Among the comments was one saying “There is a person who needs to recalibrate their sense of terror.” The response to that was this:
Although I am bearish on the future of the internet in general with AI, the concerns above just sort of made me laugh.
When it comes to doctors and lawyers, what matters are results. Even if we assume ChatGPT somehow made someone pass the bar or get a medical license – and they further had no practical exam components/residency for some reason – the ultimate proof is the real world application. Does the lawyer win their cases? Do the patients have good health outcomes? It would certainly suck to be the first few clients that prove the professionals had no skills, but that can usually be avoided by sticking to those with a positive record to begin with.
And let’s not pretend that fresh graduates who did everything legit are always going to be good at their jobs. It’s like the old joke: what do you call the person who passed medical school with a C-? “Doctor.”
The other funny thing here is the implicit assumption that a given surgeon knowing which drug to administer is better than an AI chatbot. Sure, it’s a natural assumption to make. But surgeons, doctors, and everyone in-between are constantly lobbied (read: bribed) by drug companies to use their new products instead. How many thousands of professionals started over-prescribing OxyContin after attending “all expenses paid” Purdue-funded conferences? Do you know which conferences your doctor has attended recently? Do they even attend conferences? Maybe they already use AI, eh?
Having said that, I’m not super-optimistic about ChatGPT in general. A lot of these machine-learning algorithms get their base data from publicly-available sources. Once a few of the nonsense AI get loosed in a Dead Internet scenario, there is going to be a rather sudden Ouroboros situation where ChatGPT consumes anti-ChatGPT nonsense in an infinite loop. Maybe the programmers can whitelist a few select, trustworthy sources, but that limits the scope of what ChatGPT would be able to communicate. And even in the best case scenario, doesn’t that mean tight, private control over the only unsullied datasets?
Which, if you are catering to just a few, federated groups of people anyway, maybe that is all you need.
Posted in Commentary
4 Comments
Tags: AI, ChatGDP, Dead Internet Theory, Machine Learning