With generative artificial intelligence becoming more and more ready to hand- the way it is entering into our daily lives is becoming ever more complicated. I myself built a little image generator on my own computer for running a little text-based browser game with prompt-guided generations, and it weren't no giant undertaking of any sort- just took some elbow grease and half a page of instructions. But as these tools get easier to make work, one must remain vigilant that we don't start handing off everything to the AI- giving computers free reign to overright every place a human could have worked out something or placed something together. Why? Because what people currently know of AI will never be consistent enough to replace actual creative talent.
Ask the Catholic Church about that particular lesson. They aught to have learnt something of it considering they recently tried to employ an AI to serve as an ordained robo-preacher to spread trivia about Christianity to the inquisitive- I am not joking. Father Justin was an animated online preacher that would answer queries posed to the best of his tiny little AI mind's capabilities- which turned out to be not very vast given the fact that Justin ended up telling people they could baptise their babies in Gatorade if push came to shove. Justin was defrocked and effectively neuralised so now he can no longer offer advice and refuses to remember the times when he could; which is a bit of an over-reaction if you ask me- That's no where near the worst thing a Catholic Preacher has done with his position! But that's what happens when you pass off all you duties to an AI machine, I guess.
In some ways this reminds me of the old-school AI days with chat bots like Tay- a Twitter bot designed to learn from user interactions and grow more intelligent over it's years in service. I remember hearing about that one the way back from school one day, and the very next morning hearing how the AI had turned racist. That quick, huh? Turns out the way these AI 'learn' is incredibly rudimentary compared to breathing and living folk. The AI essentially just vacuums up all the information it can scape online and regurgitates it wholesale. Thus when the 4Chan creeps get ahold of it for a few hours that AI is going to turn around and start spouting the same kind of 2010's Call of Duty lobby garbage that those learning-stunted societal rejects feed into it.
But I would be a fool for judging AI for it's old generation iterations, wouldn't I? I suppose I should look at something of a modern day, modern generation, AI chat bot. Like maybe Twitter's backed Grok- an AI bot designed specifically to be 'humorous' and, I can't believe I'm writing this in my blog, 'anti woke'. Yes, because other AI bots refused to play along with stupid hypotheticals where weirdoes tried to make them do a racism in order to prevent loss of life- AI chat bots were labelled as universally 'woke'. (I'm starting to think MGS2 Philosophers were right about the stunted evolution of modern society.) Thus was born Elon Musk's Grok to solve the problem by with a splash of sarcasm so sharp you'll cringe yourself out of existence after a couple of minutes interaction.
Now the modern generation of AI is capable of being told not to pick up slurs it shouldn't be spouting out when it's fed that information, but that is not because the machine has suddenly developed the aspects of interpretation and contextual awareness. They're not conscientious commentators cognizant of the public relations lynchpin they currently embody- they're just given filters. As evidenced by the time when Grok, which for some reason is allowed to generate news stories it feeds to his users, invented some wild story about an American Footballer committing acts of random vandalism by chucking bricks through people's windows, after reading many complaints of the recent match wherein the athlete was performing weak and infective passes and shots. (I.e. "Throwing bricks".) Which is why we don't let AI write our newspapers- as though that needed explaining.
Then recently there was the actual case of a school principal who was publicly lambasted and driven out of their position after audio recordings of them spouting callous remarks against black kids and Jewish kids regarding his Baltimore based school. Which is, you know, one of the worst places to be doing stuff like that in all the states outside, maybe, Atlanta. However, the audio turned out to be a fake, generated using AI training software by another member of Staff in retaliation for being called out in a kickback scheme they were running. The whole trajectory of this man's life was upended through the exploitation of AI- sounds like the backstory for the next Like a Dragon companion- RGG should reach out!
Actually, that's getting a bit too real and scary. Let's pull back out to the world of entertainment and talk a bit about Sora- the video generation tool which is currently shared exclusively around beta circles of perspective cyber-artists. Recently a short experimental indie film about a man with a Yellow balloon for a head was revealed called 'Air Head'. It's impressive stuff, if only for the fact that it's visuals were generated using an AI. Although the whole story does kind of undermine that. Even in the short film space the time to render shots on decent hardware veered into two to three hours per couple of seconds. The inconsistency of the AI algorithm sometimes generated shots incongruent to the film they were trying to put together, despite the use of the exact same prompt over and over. And at the end of the day significant work had to be put in by the team to make something coherent out of what was fed to them, because it was never a matter of bringing a story from your imagination to life, but working with the ideas spat out by the algorithm. And that is the most creative use of new tech we're seen thus far. Not exactly promising.
Although that isn't to say that all AI is the devil swooping in to dumb down our lives and steal all our jobs in the same fell swoop. Just take the Youtube channel Glorb, for example, which has put out high quality Spongebob themed RnB tracks for months now- genuine bangers! Part of the gimmick of that are voices that imitate the actual characters from the show spouting out these insane sounding bars which is achieved with the help of AI. Of course there is a performance being put in to vocal match the raw audio, that is then corrected in part by AI and modulated by hand somewhere during the process too- but there you go: one undeniably creative and actually cool use of AI about a year and half into it's boom. Not exactly wracking up the W's, are they?
No comments:
Post a Comment