Artificial intelligence still has some way to go

He doesn't sound 'glad all over' at the prospect

Lol

Still far from sure if this is the beginning of the end or just a passing hoax craze

https://www.independent.ie/irish-news/irish-times-removes-article-suggesting-fake-tan-is-racist-amid-suggestion-it-was-ai-generated/a289268428.html

Been doing a trawl on the AI vocal tracks on YouTube.

They range from
1. Terrible gibberish
2. Suspend disbelief, maybe
3. That's just a Michael Jackson impersonator with a backing track

Like, you could make a C90 of "Yesterday" fakeys

The AIsis one is easily the most successful.

https://foodly.tn/tips/why-do-cracker-barrel-waitress-get-stars/

ai has arrived. this page answers every question

This made some ok points:

https://www.theguardian.com/commentisfree/2023/may/15/artificial-intelligence-cynicism-technology

"To stay sane, I have had to abide by twin principles: I don’t believe it until I see it. Once I see it, I believe it."

Having said that, I think this is the bad point. You gotta look at the thing and sometimes make a judgement call that not just yes or no but maybe. People do think that it's a few notches away from improving things so that a machine could automate a job here or there. There is a lot of research being poured into automation, and even if it's a trick as long as the value of the job is bought down so that people can do it more cheaply it could be enough.

Remember when WeWork would kill commercial real estate? Crypto would abolish banks? The metaverse would end meeting people in real life?

No?

it does make some decent arguments but yeah framing it like that is silly. Crypto and Meta were obviously dumb ideas from the start which got dunked on constantly, and even still you could argue that Crypto still succeeded beyond anyone's wildest dreams, even if it did wind up leaving a lot of unfortunate rubes broke. AI on the other hand is clearly a real paradigm shift, something which may end up not replacing a ton of jobs but will clearly replace a lot of workflows. If it reminds me of anything it's the self-driving hype from 5-6 years ago, which seemed at first like it was going to revolutionize (or even destroy) the entire trucking/taxi sectors, which turned out to be a wee bit more difficult to pull off than it seemed at first. I think a lot of the concerns about AI hinge on the idea of it being able to interface well between multiple applications and codebases, and if you work in IT you'd know how incredibly hard that is. it's the reason why NFTs were such an idiotic idea, anyone who actually believed these things would allow you unique privileges/powers in a variety of games has clearly never worked on one before. which brings up another bone I have to pick with this article - "Elon Musk is clearly great at technology"...really? the guy whose code famously had to all be completely rewritten the instant he left Paypal? The guy whose cars are constantly falling apart with issues that literally no other auto manufacturer has ever had to deal with before? The guy whose dumb decision making singlehandedly caused his rocket to explode resulting in an environmental catastrophe not even a month ago?

Yeah that’s pretty much my take. I am a lot less skeptical of the transformative potential of GPT type AI than I was of crypto, the metaverse, etc. But I also think there are a lot of steps in the chain of things that would need to happen for it to really take over most “knowledge work” or whatever, and that there could be bumps or roadblocks.

It's not good enough to replace people but plenty of CEO types are dumb enough to think it is and use it as such, replacing people anyway. Already been plenty of such journalism cases.

I work in comms and am already using ChatGPT to do rough drafts and have heard plenty of anecdotal evidence of people in other professions using it, so it is already transforming things. Too early to say whether it will actually replace jobs, I suspect a bit but not a lot. It's never going to produce high-grade usable copy straight off the bat without some human intervention is my feeling.

i am half expecting the clowns at my company to try and use it to make significant amounts of training material and eventually phase us out. even though they only created our position 5 years ago.

id' be worth the layoff and ensuing poverty to watch them try and fail.

i will never use chat gpt for a first draft of anything. i don't care how good it gets. i also don't care if all my freelance work dries up because i can't make deadlines like other writers.

it is my enemy -- the enemy of humanity.

all it can see are patterns in the data. generalizations. schemata. stereotypes.

we all use these things too but as a heuristic. our categories are always complicated by our knowledge that they never quite fit. the world exceeds our understanding but it always provides us with opportunities for looking at things in a new way.

this machine has no such capacity and will not. ever. if it produces novel outputs, it arrives at them through a different means. it doesn't touch reality.

this all sounds melodramatic but it is how i feel.

i hate getting student work that was written by it, the glib competence.

we had an assembly today where some teachers talked about potential uses for gpt in our profession, like letting students use it as a first draft, or maybe using it ourselves to change the lexile levels of difficult texts. "it's here and we need to adapt" was the message.

no. never.

it's not even about being afraid it will replace jobs. i am worried about that, but my hatred of this technology is far more visceral.

I suspect you're tilting at windmills there, Treeship! No, it's never going to write a good or original novel. But it may help with some humdrum web copy, or maybe class planning for a teacher. In any case, I'm guessing it'll soon be as hard to avoid as say Google search is today.

i am afraid of a world where people are interacting with the outputs of this stuff all the time to the extent that it shapes their thinking, subtly and insidiously. like, looking at ai generated models in ads all the time doesn't seem that different than looking at airbrushed models in ads all the time. but it is different, i think. same for reading shitty, derivative web copy written by a human versus reading the same thing generated by ai. there are no cracks in the facade with this stuff. it is only derivative product, only a reworking of what came before.

i am afraid of this world and will probably never cease to be

xp I think I agree with all that, but sometimes what you actually need to produce is some derivative reworking of what came before! There's some really dry, boring marketing crap I have to write in my job sometimes, and if AI can help with that, I'm OK with it!

treeship you literally sound like a 14th century scribe who is angry about printing presses

no shade at treesh at all but i about spit my soda out, thx sleeve

I had a lot of the same conerns but they faded a bit after actually using ChatGPT for an hour and seeing how sterile and boring everything it spits out is

i started the Talk to ChatGPT thread and got bored of using it in less than a month

I haven't seen chat gpt produce something enjoyable to read, but it can do a fair job of organizing a set of decent points into a basic 8th grade level essay. It's hard for me to see that replacing any type of writing that people seek out because they enjoy reading. Maybe someone will figure out how to make it churn out something that is stylistically pleasing and original, who knows?

Yeah I don't really think the soul of humanity will be lost if we let a robot do the first draft of, like, a hoover instruction manual, but I do find treeship's hardline stance here very romantic.

Ah, to be young.

Remember when WeWork would kill commercial real estate? Crypto would abolish banks? The metaverse would end meeting people in real life?

No?

― longtime caller, first time listener (man alive), Monday, 15 May 2023 bookmarkflaglink

I remember some marketing like this. One or two countries were playing with using crypto as currency and the like.

Meta is definitely sold as 'no need to meet people in real life'.

I don't remember the WeWork hype.

Yeah I don't really think the soul of humanity will be lost if we let a robot do the first draft of, like, a hoover instruction manual, but I do find treeship's hardline stance here very romantic.

chatGPT is becoming a real engineer pic.twitter.com/3aTXYgLki2

— gaut (@0xgaut) May 15, 2023

hahahaha amazing

Lol, there is a lot of goofiness with Chat GPT and the various AI stuff that has come around in the past couple years, but I also think that this is much more real than Meta or cryptocurrency. I've been keeping an eye on the image generation stuff during this time and the rate at which it has progressed is truly stunning. I think of this all as the next generation in computing as opposed to a scary thing that will replace humanity. It's going to result in significant changes, just like microprocessors, the internet, and smart phones did, but it is not going to be the end of creativity. More likely it will be used as a way to enhance creativity. None of this stuff can actually used in a creative way if the people guiding it are not themselves creative as well, it all comes down the human mind behind it.

i want to believe that but i am skeptical. the thing itself is a mind -- a hive mind. hito steyerl's piece in the new left review is so critical i think:

Mean images are far from random hallucinations. They are predictable products of data populism. They pick up on latent social patterns that encode conflicting significations as vector coordinates. They visualize real existing social attitudes that align the common with lower-class status, mediocrity and nasty behaviour. They are after-images, burnt into screens and retinas long after their source has been erased. They perform a psychoanalysis without either psyche or analysis for an age of automation in which production is augmented by wholesale fabrication. Mean images are social dreams without sleep, processing society’s irrational functions to their logical conclusions. They are documentary expressions of society’s views of itself, seized through the chaotic capture and large-scale kidnapping of data. And they rely on vast infrastructures of polluting hardware and menial and disenfranchised labour, exploiting political conflict as a resource.

That is more a piece of writing/performance than the reality of the situation. It's shameful to get a piece published with a term like 'data populism' in it.

why? it seems descriptive of a real phenomena to me

"It's going to result in significant changes, just like microprocessors, the internet, and smart phones did, but it is not going to be the end of creativity."

All of that stuff might create chaos, but only with significant human input.

why? it seems descriptive of a real phenomena to me

― z_tbd, Tuesday, 16 May 2023 bookmarkflaglink

So all the bits you may recognise haven't been independently created by AI. We already have "vast structures of polluting hardware", for example.

i haven't read the new left review piece. when it comes to the use of the term "data populism", i'm going off the first two sentences above:

"Mean images are far from random hallucinations. They are predictable products of data populism."

that sounds right to me. data populism, to me, is related to recreating the average of what vast swaths of people "like", which, imo, leads to bland crappy art. i've typed too many words in the past about all of that, but that's because it's a real phenomena, both in AI and non-AI (like the MCU)

the stuff in the quote about "social dreams without sleep", "processing society's irrational functions to their logical conclusions" -- i wish that were more true, or that i felt it to be more true. that would be more interesting than art that looks like deviantart. it's funny to me that everyone complains about how dall-e can't do hands and makes too many fingers and stuff -- that is actually interesting! it's gonna suck when it makes normal hands and fingers!

It will make normal hands and fingers, so what?

there is a memetic quality to discourse. when people sit down to write or even just have a conversation, they aren't saying new things every time. they often don't know where their influences are coming from either. as barthes said, "the text is a tissue of quotations drawn from the innumerable centers of culture." (this is true of other types of creative work too, like the production and dissemination of images is also a kind of discuourse).

however, it is one thing to recognize this "automatic" quality that is present in communication. it is another to fully automate it. this will just harden all of our patterns and assumptions. it's like, with these machines culture can reproduce itself without being mediated by thinking at all.

i just don't think this can possibly be good.

with these machines culture can reproduce itself without being mediated by thinking at all.

entire post was otm

treeship otm. this 'technology' isn't going away because it will be extremely useful as a concealment. the more i think about it the more the cheap captive labor that drives it is the key. i also am viscerally repulsed by its output and can't entertain anyone finding any merit at all therein. i've seen, ignored and been happier for ignoring trends before this and i'm sure this won't be the last.

would love to see an AI bot in a pitch meeting

technology has already shifted our culture a lot, even subtle things like spell check on phones I think changed the way we communicate with each other and made it a bit less personal. Twitter and YouTube have spun off their own weird brands of comedy which have materialized in sketch shows and stand up. it wasn't all that long ago that referencing a meme in real life was a really weird thing to do. I guess its natural to be afraid of the wider implications of this but I think it ain't all bad. My kids growing up in a world where you can literally just ask your watch a question and get the real answer is kind of a cool thing, if I'd grown up with that I think it would've rewired my brain in a few helpful ways

this whole thing reminds me a bit of these old ideas of how say the drum machine was gonna put drummers out of business, or how the Fairlight and Synclaiver were gonna kill the orchestra, and people spent a quarter million dollars to produce albums like Shout! and Pulling Rabbits Out of a Hat which sounded like shit. and that's not just revisionist history, people thought so at the time too. it's the same with a lot of AI stuff, yes the images they generate now look amazing but they're also soulless and completely uninspiring. the writing is technically impressive but also boring as hell. but like with drum machines for instance you had these folks who realized the real potential in them wasn't just replicating existing music but in doing stuff humans couldn't do, and that is where a lot of really great stuff lies.

so I think the shit like "AI's gonna generate an infinite amount of Drake songs" is just going to be a dead end, no one outside of blue checkmark types are gonna think that's interesting or gamechanging at all. but there are a lot of cool things you can do now with AI that simply wasn't possible before. some of the deepfake stuff on YouTube is really clever and hilarious, but the common thread is that all the good stuff is something that's been thought through and edited by a real person. even like that Twitch stream that did endless Seinfeld or Steamed Hams is a fascinating art project, because it's not AI replacing art, it's a new kind of art that's funny and bizarre because you know it's AI. that's the kind of thing I expect to see a lot of in the future.

My problem is there are a number of legitimate concerns that you could reasonably discuss about it but the issues that everyone focuses on are WHOGIVESASHIT.jpg

Just like when auto-tune gained took off in the 21st century and a valid gripe was "it makes a lot of pop singers sound sterile and homogeneous" but all we ever heard was the meritocratic argument that it was making bad singers into amazing ones, as if a tool that could fix wobbly pitch or create a neat robot sound was turning Lil Wayne into Mario Lanza

To me that's the exact same story, we focus on the shittiest derivative and unimaginative uses of a thing, which doesn't really make a whole lot of sense because 99.9% of attempts to be creative in every medium ever have been shitty, derivative, and unimaginative.

Sen. Hawley asks whether A.I. is going to be more like the printing press (good) or the atom bomb (bad).

— Scott Nover (@ScottNover) May 16, 2023