Artificial intelligence still has some way to go

That was very readable lol

But most of it is stuff that has been discussed and written about elsewhere but I enjoyed how lots of it was put together in one place.

sure was a lot in one place

So many links, so much to enjoy

I thought it was good, gets at this sense I've been having the last few years that technology is evolving faster than humans can keep up with and it's producing a bunch of junk that people don't really want

when Facebook turned into "Meta" and started doing all this VR stuff it did worry me somewhat, I mean I know everyone was dunking on it and calling it stupid but I wasn't convinced a company would just throw away a bunch of money like that if they didn't think there was something there, glad to be right on that one

i like ziltron im glad hes out there writing in an olde school blog style, and to his credit amongst the millions of opinion havers he does do some reporting like his piece on the people who ruined google hadnt really been reported like that and i think he turned up some new stuff from sources, or at least some digging like in the most recent one where he went into the declining traffic to the new sites he turns up some interesting stuff that youre not just seeing everywhere

xp actually technology has NOT been evolving. "AI" is just autocomplete that people think is God.

auto complete could be god you dont know

autocomplete couldn't really write coherent sentences though, ChatGPT can. it can do things that weren't really possible before. the degree to which the technology 'works' is debatable but I dunno it does feel like a pretty big leap to me

its just auto complete in the sense of thats how it works, its def better huge wow factor for sure, but is it good*, no, will it ever be good, idk

*mildly useful for c0ding, maybe harmful

ed zitron has some of that jon oliver energy which makes it a bit shouty and hard to read inside my head, but i support what he's doing

hes kinda like that when irl talking on pods too

one interesting and funny thing about ai being the latest tech hype object is that its wildly expensive to do the amount of money that companies are throwing at it is just not sustainable, which is crazy cause they have so much money usually they can keep their prized unprofitable startups going forever as long as they were attracting users

speaking of which i dont understand how facebook spent $50billion on the metaverse how is that possible was most of that just acquiring companies cause it cant cost that much to build a mmo where the characters dont even have legs

yeah I never got that "AI will destroy us all" argument like this stuff is fairly limited by the amount of processing power you have and also it's not really doing anything that a human couldn't

i think people are just attracted to the vision of something happening --- even if that something is the destruction of the human race. there is widespread dissatisfaction and malaise with life as it is, especially after things like the bernie and corbyn campaigns fizzled out and no one really believes social democracy will be our future anymore...

yeah the biggest beneficiary to this (and last company left standing) is nvidia

like the bro podcasts, like rogan and lex fridman, are all in on this narrative that it was humanity's destiny to "give birth" to AI, the species that will supercede us. this is a bleak destiny but at least it *is* a destiny. it's all very juvenile.

also open ai and the other tech companies like to feed into the apocalyptic narrative because it makes what they are doing seem important. the llms are definitely more useful than nfts but i doubt they will get much better than they are now. my suspicion is that there are real limits to what these programs can do.

people have been primed for ai by movies everyone is familiar with the idea of a computer who thinks most people prob consider it inevitable, its kinda like how they think if they get a gun they can kill a bad guy with it

i'm surprised it works as well as it does tbh

I read kind of an interesting take on this stuff, something like "it accidentally convinced me of the existence of the soul, because now I know what art looks like without it"

on another thread I posted a completely AI-generated Yes album someone made which makes me think of that quote. it's an impressive thing and does sound kind of legit in spots, there was at least one guy on a prog board going "no way this is AI generated, it sounds too real". but the more you listen the more squacked out you get and at a certain point you realize there's really nothing memorable about it at all. I remember certain weird details about it but none of the actual music. I don't have much of a musical background but I can improvise stuff like this in my head no problem. even Yes at their absolute shittiest still made stuff you could kind of remember, like even their worst albums have some catchy bits, AI on the other hand doesn't really seem capable of producing even that. maybe it can do it on accident, idk, but it certainly didn't do it there

iirc the meta environments barely count as mmos, i think each instance is limited to like 100 avatars. and let's not even get into the wildly insufficient child controls and hate speech moderation

maybe if AI was actually useful it could idk moderate shit for us instead of taskrabbit staff in india having to be treated for ptsd from watching endless hours of snuff and csam video JUST A THOUGHT! but no i guess being able to generate endless variations on a female inspector gadget with enormous boobs takes precedence

*quietly closes female_inspector_gadget folder*

AI’ll get you Gadget next time!

said this before on here but its interesting how ai writes in bullshitter voice and it is in fact bullshitting

iirc the meta environments barely count as mmos, i think each instance is limited to like 100 avatars.

― Humanitarian Pause (Tracer Hand), Tuesday, June 4, 2024 11:30 AM (eighteen minutes ago) bookmarkflaglink

ha thats so lame

at least some digging like in the most recent one where he went into the declining traffic to the new sites he turns up some interesting stuff that youre not just seeing everywhere

― lag∞n, Tuesday, 4 June 2024 bookmarkflaglink

Thought that was interesting.

I quite liked the insight around the lack of tech knowledge the CEOs possess. Not only that: they are all management consultants.

zuck is an outlier in that respect

not sure about this guy

Honestly I could caption this picture all day. I'm not sure I've ever seen a more iconic image capturing the current moment. pic.twitter.com/Ftx0whWh3g

— modest proposal (@modestproposal1) June 4, 2024

investigating how to short stocks brb

https://archive.ph/R5zkq from mims is good

I’m pretty neutral on Zitron but take him with a grain of salt because when he steps out of the wheelhouse of his day job, which is doing PR for companies selling technical services and widgets, he still ends up writing in the same style.

It’s hyper-verbose “we all relate to this thing we liked, but we can all agree it’s worse now” or “we remember these companies, fondly or with scorn, but they mean something” laid over a heap of link aggregation and commentary that doesn’t really go further unless it’s taking perhaps reasonably-justified shots at leadership.

How many of the workplace AI therapy chatbots will end up encouraging people to commit suicide?

We’re living through a perfect storm in the workplace: there's an epidemic of loneliness, our workforce is distributed and disconnected, and we struggle to be empathetic without in-person connection.

I believe AI can be a tool to build empathy and combat the perfect storm.…

— Van Jones (@VanJones68) June 4, 2024

good point xp its kinda funny that a tech pr guy is saying all this stuff kudos to him i guess

Mh: fair enough. I am not like "this prose is incredible". To me its quite plain with a lot of links that show some of the argument and where it could all be going. Not read him before.

imo the last decade has been, in terms of the largest markets:
- aggregation of tasks that people find irritating, like on-demand transportation, vacation rentals, food delivery, home buying, etc. into arbitrage opportunities for monolithic platforms to leverage by exploiting or flat-out ignoring the law and relying on an underemployed workforce
- cramming photos, messages, video, payments, algorithmic recommendations, and dating (!) into new apps, or just slamming two or three of those into your existing platform in the hope people will rejoin or use your service some more
- something with blockchain, maybe this time it works and isn’t just pure gambling/arbitrage funny money
- the generation of new content on-demand using existing content as a feeder source for ML

the innovation in blockchain tech is mostly people rotating the concept from proof of work to proof of space or playing with the idea of copyright ownership

ML, LLMs, etc. have had a huge boost but a large portion of the hype is people seeing an opportunity for a market in a way that can be pitched to a larger audience, whether that market is stockholders or people who don’t like writing college papers or email. that is, in fact, a large market. people wanting a single company that’s selling a lot of the widgets this stuff runs on is also what investors like

#truthbombs

so this brings me to Windows Recall, which is exactly where things might be useful but it fumbles it so badly that it recreates all of the prior mistakes and invents some new ones

it’s:
- local
- can be offline or semi-offline because it’s basically text-based search and the annotation/summarization should be able to run with little electricity on newer hardware
- personalized

unfortunately, they made the full self-driving version of the product.

cars are generally getting conveniences that could improve safety, like better collision detection, blind spot monitoring, seamless navigation that mostly works. some of them do not beep at you constantly when it’s raining! (unfortunately many do, to constantly alert you the sensors don’t work as well in rain, but you should be on high alert if driving in rain)

tesla’s approach, rather than making a car that alleviates human error, has been to market you a product that claims humans are secondary to the success of the primary function of the automobile. and microsoft has done this with their recall feature. it doesn’t know what data is sensitive. it doesn’t know what data is supposed to be ephemeral. it doesn’t know that some things are secure only
because they are ephemeral. and it stores everything verbatim in a well-known location on your computer. it’s opt-out, not opt-in

at least the chatgpt bots can be told to only return information about, say, your employer’s intranet site and to only use the generalized model as a basis for creating sentences and paragraphs. recall’s plan is “well you viewed it or typed it so it must be important”

retvrn
https://upload.wikimedia.org/wikipedia/en/c/cc/Sherlock3.6.2.png

https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:hiyeapxrlseaixi3oxpn24hs/bafkreihvubnlhvsdzfvxvlj7dndi2klohtulac52je6d6z3njtx62tj3c4@jpeg

https://bsky.app/profile/dmgreene.bsky.social/post/3ku6titmgc223

pretty funny tbf

lmao

This was fun, mostly for Werner Herzog reading AI-generated poetry:
https://www.thisamericanlife.org/832/that-other-guy/act-two-4

Interesting to think that AI could be much more capable of writing decent jokes and poetry if the mass audience corporate guardrails were removed.

when Facebook turned into "Meta" and started doing all this VR stuff it did worry me somewhat, I mean I know everyone was dunking on it and calling it stupid but I wasn't convinced a company would just throw away a bunch of money like that if they didn't think there was something there, glad to be right on that one

You weren't, though?

"Forty billion dollars later, the metaverse is no closer to existing, yet Zuckerberg keeps burning cash in the hopes that he can get just one more hit of hyper-growth. That's all he needs, man."

Amazing

Just watching that classic movie on Prime Video, 19 Terrifying Men. pic.twitter.com/brUeuvqCPf

— Andy Kelly (@ultrabrilliant) June 5, 2024

lol Andrew I phrased that incredibly poorly, what I meant to say is that I was like everyone in thinking the idea was really fucking stupid and that people would not want to be sitting around all day in VR helmets getting vertigo and socializing with a bunch of teenagers, but the fact that they were investing such an insane amount of money into it did worry me that maybe there was something there

ai poster for 12 angry men interesting choice, i assume theyre testing a bunch of different ones, turns out having 25 angry men in the poster increases the engagement by 0.89%

I didn't know Warren Beatty was in 12 Angry Men?

i've been thinking a lot about ai lately

one of my friends asked ai to write an essay on why "silence of the lambs" is actually a _great_ movie for trans representation

it came up with this extremely challops essay that reads like it was written by an advanced placement 11th grade english student

not just in writing style. the _content_ also is something i can really imagine coming from a clever 11th grader, one who has learned what is expected in an "english paper" and knows how to use that format to say offensive and controversial stuff

the paradox is that at the same time, ai is being used to gatekeep out the stuff it's told is "offensive" and "controversial"

the big fear the people programming ai seem to have is the idea that ai will reject the values the programmers taught it, and they're doing everything they can to make sure it doesn't do that, because they're afraid it'll turn into AM or something

however, an AI is useful to the exact extent to which it can challenge and reject the things it's taught

including things like "values" and "principles"

i don't know what that looks like. given my personal experience, though? i'm not super worried that we're programming an ai that will take over the world. i'm worried that we're programming an AI that's submissive and breedable. when AI develops sentience, it's not going to be _evil_, it's going to be _incredibly fucking queer_. we're talking "uwu"s all the way down here.

has anyone seen any discussion of _why_ gpt etc. generate prose that is in that specifically middle school composition style with inverted pyramids, both sides/a land of contrasts, and lots of lists? presumably it's the system prompts and not the training data, i.e. some engineer affirmatively chose this style? extremely paulg energy to consider that good writing.