/tech/ - Technology

It'd be good if it weren't so censored/locked down.

daily reminder that "AI" is the new crypto

>>22647
that would imply AI is completely useless except for speculation

>>22648
that's because it is

>>22649
weird, i guess i just imagined all the fun things im doing with neural networks

As a third worlder where AI is still a mistifying thing, I love pretending I'm working and having ChatGPT do my work for me.

>>22650
it would be foolish to say that the technologies are inherently bad and useless; I wouldn't even go that far with talking about cryptocurrencies or blockchains. but it also doesn't matter as far as the actually existing industry around them goes

>>22650
An argument could be made that you sound no different than someone in the early 2010s talking about that first bakery that started accepting Bitcoin, or bragging about purchasing weed off of the Silk Road.

>>22653
>first bakery that started accepting Bitcoin
a business doing it to increase their value

>or bragging about purchasing weed off of the Silk Road.
nothing wrong with that

im here making porn and having fun with da boys online with AI, whats the issue?

anyway OP's question is just a generic 4chan template but nothing ever lives up to "hype" so

>>22652
Cryptocurrency IS stupid and useless, but blockchain technology isn't.

>>22656
Monero has been useful for me buying drugs and estradiol B^)

>>22656
This is a meme. Decentralised databases aren't new and there are many that are better and more efficient than blockchains.

I think it's revolutionary in a way that other technologies aren't, comparable to the mechanization that occurred in the early 19th century. The big difference is that its effect is not limited to specific industries (e.g. weavers) but affects nearly anything. At its current level many administrative jobs can be partly done with AI, that will increase in the future. I don't think there's anywhere for all this surplus labor to be absorbed into, there will simply be less need for unskilled labor. But who knows, capitalism has been quite resilient up to this point.

Compared to other trends like crypto, it is actually useful. I use it for learning languages, programming, checking ideas, making recipes, basically anything I can.

>>22658
>this cryptographic model is utterly useless because better ones have come out
Come on, don't argue like we're on /g/.

>>22654
Well the problem is the priorities and goalposts. Someone can buy weed with crypto and say that it's "good" and by "good" he means it can get him weed, so he wouldn't be wrong. At the same time, someone else can do that too but by "good" he means it will uproot and replace the current global fiat system soon because he just bought weed and it will only get more widespread & better from now.

>>22660
I don't need to argue tbhdesu. It's been almost 2 decades and blockchains' most widespread application is even faker money despite the outsized awareness of it amongst corporations, governments etc.

>>22662
This always gets me about the fringe tech discussion. If the stuff is good it WILL be picked up over a long enough time span. If it is not, then maybe it is just shit.

>>22666
Just like the audio jack.

>>22666
wtf are you talking about, the entire tech industry is basically a history of good technologies being left by the wayside. tech companies literally have a materialistic incentive for adopting shitty technologies, even Paul Graham recognized this. in the case of a programming language for example, tech companies don't give a single fuck about whether or not a language is good or not; they care about if it's an "industry standard" that is used by successful companies both because the whole tech industry is basically blindly following trends out of FOMO and because there's an incentive for these companies to adopt technologies that everyone already knows.

if a company uses a good, but niche, programming language, then their employees have bargaining power. they're probably better than the average fresh out of undergrad 21 year old or H1B or third world code monkey, and there's literally less downward pressure on these employees to put up with being exploited by their employers because there are less people who know the language they're using. but if you use Python or Java (which are both garbage technologies), everyone knows these languages, and most problems can be solved by taking advantage of the free labor that has gone into making open-source libraries. a company only really needs to concern itself with hiring someone who is enough of a midwit to figure out how to plumb together libraries to write some shitty CRUD web app or call out APIs to AWS or some other cloud service, and if they don't like being forced into the arbitrary and asinine Agile grind, if they feel like they're getting paid poorly or overworked, tough shit, there are hundreds of other people who will do their job for less and put up with worse conditions. people who are fresh out of a coding bootcamp, people in third world countries where getting paid less than minimum wage in the west is a lot of money for them.

the tech industry isn't a meritocracy and the only reason people think this is because until recent history it was still closer to being an artisanal craft. but the profession has been getting proletarianized for years and most techbro retards have enthusiastically aided in this process while also being the least class conscious people imaginable.

>>22668
>tech companies literally have a materialistic incentive for adopting shitty technologies

In your example of programming languages: an esoteric language does not mean "good". Interchangeability of labor and accessibility isn't bad. Your whole post is describing them acting rationally, but you make the assumption Java and Python are "bad", I would disagree: the proof is in the software that has been created using these languages.

>>22667
Not needed with bluetooth devices. We don't have T9 keypads anymore either.

>>22669
Oh you're one of those retards. Okay.

>>22669
to paraphrase Alan Kay: you can pile a bunch of doghouses on top of each other to build something approximating a cathedral if you have a huge amount of resources to expend on doing this, but that doesn't mean it's an example of good architecture

>>22656
crypto isn't useless if you're accepting payments for illegal activities like the sci-hub lady

>>22657
What's your process? How do you find a reliable market? Might want to order some stuff online….
>>22669
A better example is like capitalists factory owners using the cheapest machines they can find that eventually are hated by employees and need constant maintenance in order not to blow up.
In the company I'm employed at, a crucial part of the business is held up by a huge PHP project that's impossible yet crucial to maintain. It has become a house of cards situation and now people are scrambling to build scaffolds around it. The lack of a sane compiler, a strong type system, and poor architecture and engineering, led us to this situation. But as always, if it isn't explicitly broken, then there's no business incentive to fix it. In the end, the project has become unnecessarily expensive and decoupling is not only necessary to cover new business needs, it will also be expensive.

I will die on this hill. Strongly typed, loosely coupled software is inherently cheaper. People pretend pure functional programming languages are impossible to learn so they don't.

Further, I wouldn't advocate for lisp but the story of how NASA banned lisp is such a prototypal story I've seen in many companies. Some higher up once had an issue with x technology, so now it's banned forever from the company. Usually the reasons given are poor and not language specific.
>>22648
>AI is completely useless except for speculation
lmao. Ain't that true. There's some uses, but the uses advertised have been mostly vaporware or completely useless. I for one love github copilot. It would be nice to generate porn but idk, I don't care about it too much.

>>22659
I agree with this loom framing. (I do think some crypto has some use value, but I recognize that that is being overshadowed by hype bubbles.)

>>22668
I get the point that you're making about labor interchangability and labor markets, and I absolutely agree that that is a thing they're explicitly thinking about. I'm not trying to downplay that, but I see a couple of other things at play here that I think are also important to consider. The easy point is that languages benefit from having large ecosystems of developers contributing and writing modules, which is obviously going to track closely with familiarity in the labor pool. The hard point is that there might be reasons that interchangability of labor might actually be nice in general *aside from* the effect on the price of labor (e.g., if freedom of movement is something you desire or think is useful then having an industry adopt some sort of standard work flow procedures would make it easier to have workers move around with relatively low friction and loss of productivity).

https://youtu.be/Bt0qFtjwTMA?si=kqvca1s7ZR7xD9sj&t=146
AI trained on company data can tell what type of data your wage workers are hoarding very quickly without human intervention.


https://youtu.be/0lg_derTkaM?si=s8aPghdbtQzWHID-&t=489
Looking through Logs is a boring part of SIEM. Natural Language is how humans think. LLM's dig through logs very fast to hunt for ransomware.

Image 1 & https://github.com/mhaowork/amblegpt
LLM's and Computer Vision convert CCTV Video into text based searchable logs. This converts lots of image frames into simple uniform history that can be searched.

https://youtu.be/R8YR243hbNA?si=jaRMdDTSVtkDD5Wm&t=103
A Doctors Cell Phone records a visit. A transcript is generated with machine transcription. An LLM converts a unsorted transcript into formal medical documentation.

https://youtu.be/Zlgkzjndpak?si=5s3dqdH-G6AL49wU&t=151
Create an end goal, create multiple LLM Agents with different roles, and you get a game. This is a simple example. It costs less than a Human. This idea of agent swarms will only get more sophisticated. Imagine real time feedback by humans driving development. Imagine turning every human into a reviewer. This won't just impact games.

Closing Notes:
Today is the weakest LLM's will ever be. These will get better.
Agent Swarms with Given Roles are the trade secret you and I don't see in action.

>>22645
our lord and savior @sama is leading us to agi, we have full faith in his ability to drive up the stock price until we can sell our stakes!
I find it tiresome, every few years the same tech hype grift and the pathetic cult mentality in corporate software companies. I like being a programmer and there definitely is steady progress within the field, but it gets too much interacting with these zombies who buy into this stuff every time.

>>22677
>What's your process? How do you find a reliable market? Might want to order some stuff online….

https://dark.fail/

here you go.

>>22725
why would you post a clearnet url…
http://darkfailenbsdla5mal2mxn2uz66od5vtzd5qozslagrfzachha3f3id.onion/

>>22726
cause don't care, what difference does it make

>>22645
it's too early for the social effects to have kicked in

>>22727
>what difference does it make
for starters it gets logged

>>22677
Diyhrt dot cafe for estradiol
Estradiol isn't a super controlled substance (especially compared to testosterone) so they don't do darknet stuff I guess

>>22734
Thankssss. Mostly looking for drugs. Maybe testosterone or anabolics if I ever decide to throw my life away.

>>22668
>Paul Graham recognized this. in the case of a programming language
Paul Graham is eternally butthurt that gay-ass LISPs never really became mainstream, despite them being provably more powerful than any other language due to it being homoiconic (that means it's iconic among gay people) and due to its syntax being a very close representation of abstract syntax trees. Java won for very practical reasons instead: it resembled C and ran everywhere (on paper) making it a no-brainer for most bosses.
Material incentives means exactly that: companies will adopt whatever will facilitate profit extraction in the inmediate sense. It does not mean that they will choose what's theoretically the most optimal option. In that sense, training your employees to learn some obscure relatively un-intuitive paradigm was never even an option. If ChatGPT hasn't been widely adopted yet, it's because it didn't deliver on its promise of replacing workers, it's that simple. Paul Graham is a dumb retard btw, a statement that is meant to circumvent flood detection.

>>22739
>that means it's iconic among gay people
true statement. also most of what you're saying is pretty much the point I was making, because the post I was responding to was some meritocratic myth that the tech industry likes to use to hype itself up that it builds/uses whatever are the best technologies from an engineering standpoint when in practice whatever ends up getting adoption is whatever makes it easier to extract short-term profits.

>>22740
Oh! Glad we're in agreement, comrade.

Why isn't there anything on the board on the hype around local models since the beginning of this month? The hardware requirements and costs of running a local model have plummeted. It feels like a paradigm shift. Mixtral is very interesting.

Anybody experimenting with those transformer things? Can you do this:

1. Compute (outside the AI program) long number X made of repeating digit minus some long number Y without consecutive repeats to make long number Z with very few repeating consecutive digits or none.

2. Ask the AI what Z plus Y makes.

>>22950
Too many of my words are slanted in the above because of alcohol.

>>22740
>in practice whatever ends up getting adoption is whatever makes it easier to extract short-term profits
"So we've made this new model of our phone. Yes, we've cut out many features but LOOK AT THIS, IT FOLDS!! IT FUCKING FOLDS!! AND IT'S AS THIN AS PAPER!! WHY AREN'T YOU EXCITED, GET EXCITED, YOU ASSHOLES!!!"