Sick of "AI"

[joshfrog]

Hi There

 

This is more of a rant then anything but is anyone else sick and tired of always hearing about "AI" stuff, new devices, new software, new tools for "AI".

I watch WAN show, TechLinked, GameLinked and other youtube channels as well as read other tech news sites that just keep talking about "AI" crap.

 

I'm just so damn sick and tired of it all, I really couldn't care any less about it.

Posted two poll's just curious what people think.

 

Thanks All

[Birblover12]

It's a marketing term, exact same thing as sticking "GAMING!!!" onto any product along with some cheap LEDs and charging 3X the price because it creates the illusion of something which is more than it actually is. Companies love to play follow the leader and follow whatever the largest companies are pushing as the "newest thing". 

[Kisai]

I think the hype around it it's largely over-promising.

 

The most post practical uses of AI is in assistive technologies (eg auto-captioning, text-to-speech, machine vision identification, early-warning systems in safety/weather/disaster-prevention systems)

 

But the AI we keep hearing about is generative AI. A type of AI that has no practical use, and only works by being trained on copyrighted materials, without any supervision or curation of the data ingress.

 

Generative AI is going nowhere. It's used for deepfakes, and there will eventually come a point that regulations come down the pipe and kill all viable commercial use of generative AI. This will either come as regulations on what content the AI can consume or liability for the content it generates, as well as the need to clearly identify materials that aren't AI generated from AI generated due to the erosion of trust caused by AI use in journalism.

 

Like the idea that a car will learn how to drive using cameras and AI alone is pretty foolish. What needs to come down the pipe for that is municipal traffic control systems (eg traffic lights, signs, and temporary diversions) being communicated to the car in real time, as the primary traffic control scheme, and using something similar to CBTC in train signaling to actually control the car spacing. This leaves the car and the driver only in control when municipal control is not available, and the cameras control used when there is a conflict between municipal control and driver actions. (Eg cameras indicate an obstacle when the municipal control indicates the lights are green.)

 

A lot of the over-hype again, is with the idea that AI might replace creatives. This is not going to happen, nobody is going to pay for machine-generated works that they can produce themselves. AI can't predict the future, it can't imagine, it can only hallucinate, and that hallucination is largely a product of it consuming poisonous discourse from the internet and losing it's own context. 

 

[wasab]

AI is just buzzwords. since chatgpt, everyone has always been ai this and ai that. it is way overhype. 

[12345678]

8 hours ago, wasab said:

AI is just buzzwords. since chatgpt, everyone has always been ai this and ai that. it is way overhype. 

Even before, with all ai photo assist, ai writings and so on,

 

Currently is going more mainstream, and as such I am sick of people endlessly talking about stuff they have no knowledge of like if it was the newest thing, or oligopolies making pr moves by replicating old as fuck stuff as it was theirs 

 

Like at my workplace, somehow w11 is news, but fuck me it has been around for at least 3 years 

[starsmine]

In all honesty, AI will be like the internet. The tool is overly blunt today sure. Hammer and nails and all that. But just like how much and how rapidly things changed in the information age. AI will in the next 20 years. Internet beginning in the late 80s/early 90s did not really hit its stride until after the dot com burst. 

Saying im sick of the hype just seems misleading. Pandora's box is open on this one. 

[wasab]

I use copilot. Useful to save myself some time. Chatgpt is overrated tho

[thrasher_565]

ai been around a long time... dont no the first thing about ai but like there were ai bots on msn...

the big bad thing about ai is that it make robots or programs to repalce people jobs. and well wepons. but thats always been at the for front is develiping for wepons and maybe it ill be in your house at some point.

 

but then there things like why is there no ai that fixsis by spelling yet...

 

posably dell had the first so called ai to play chees posably, i no there was a way to send messages before then way before then internet.

Edited January 20 by thrasher_565

[Monkey Dust]

I think AI may destroy the internet, at least the www bit. My hypothesis is that first vast amounts of AI generated content will crowd out all other content. And the as the new content begins to be generated based on the AI hallucinations that went before it, it will become near impossible to tell fact from fiction. In time, we will just give up on the web.  

 

I think the hype right now isn't so much overdone, as premature. It's just a very large statistical model, it doesn't really understand anything, which limits its usefulness somewhat.  

[NadiaMayer]

On 1/18/2024 at 6:34 PM, Kisai said:

I think the hype around it it's largely over-promising.

 

The most post practical uses of AI is in assistive technologies (eg auto-captioning, text-to-speech, machine vision identification, early-warning systems in safety/weather/disaster-prevention systems)

 

But the AI we keep hearing about is generative AI. A type of AI that has no practical use, and only works by being trained on copyrighted materials, without any supervision or curation of the data ingress.

 

Generative AI is going nowhere. It's used for deepfakes, and there will eventually come a point that regulations come down the pipe and kill all viable commercial use of generative AI. This will either come as regulations on what content the AI can consume or liability for the content it generates, as well as the need to clearly identify materials that aren't AI generated from AI generated due to the erosion of trust caused by AI use in journalism.

 

Like the idea that a car will learn how to drive using cameras and AI alone is pretty foolish. What needs to come down the pipe for that is municipal traffic control systems (eg traffic lights, signs, and temporary diversions) being communicated to the car in real time, as the primary traffic control scheme, and using something similar to CBTC in train signaling to actually control the car spacing. This leaves the car and the driver only in control when municipal control is not available, and the cameras control used when there is a conflict between municipal control and driver actions. (Eg cameras indicate an obstacle when the municipal control indicates the lights are green.)

 

A lot of the over-hype again, is with the idea that AI might replace creatives. This is not going to happen, nobody is going to pay for machine-generated works that they can produce themselves. AI can't predict the future, it can't imagine, it can only hallucinate, and that hallucination is largely a product of it consuming poisonous discourse from the internet and losing it's own context. 

 

i dont think commercial AI going to be killed by copyright from an Ironic paradox

Disney or WB and such would be the ones who have the power and lawyers to kill stuff like LLM and dally like models but in a conflict they also are one to benefit the most from said tools as they cut down work forces.

so yeah things are gonna be interesting  

or Disney does what Disney to the public domain and uses it while knock out the lower ladder rungs for upstarts

[Lunar River]

1 hour ago, NadiaMayer said:

Disney or WB and such would be the ones who have the power and lawyers to kill stuff like LLM and dally like models but in a conflict they also are one to benefit the most from said tools as they cut down work forces.

and do what to them?

 

the models are everywhere, people have them and are using them. they can mess with the online sites that allow you to generate stuff on their servers, but they can't do anything about local models

[NadiaMayer]

7 hours ago, Lunar River said:

and do what to them?

 

the models are everywhere, people have them and are using them. they can mess with the online sites that allow you to generate stuff on their servers, but they can't do anything about local models

That is  what I'm saying the genie is out of the bottle and it's useful I can not see anything that gonna stop it the close we got is SAG-AFTRA they already are slipping.

while LLM and ai art models make piss poor work they still major make lotta work easy and speed up producton with minimal loss (assumed they still keep humans in the loop) the tools are simply too good. ceos care about money and from my view lowering the workforces is better than letting few models slip pass.

if anything they gonna do both make local models near impossible do legally while hording the tech form themselves using some BS loophole.

[NobleGamer]

On 1/18/2024 at 3:04 AM, Kisai said:

But the AI we keep hearing about is generative AI. A type of AI that has no practical use, and only works by being trained on copyrighted materials, without any supervision or curation of the data ingress.

 

Generative AI is going nowhere. It's used for deepfakes, and there will eventually come a point that regulations come down the pipe and kill all viable commercial use of generative AI. This will either come as regulations on what content the AI can consume or liability for the content it generates, as well as the need to clearly identify materials that aren't AI generated from AI generated due to the erosion of trust caused by AI use in journalism.

 

Is there no use in being able to search for or ask a website something, and instead of getting a list of search results, you get the content you need without having to look through the search result or have your answer buried in a page's ads/fluff/formatting?

 

Because that's generative AI's most basic summary of what it can do: It takes Google's "I'm feeling lucky" button and tries to isolate what is loaded to only what you need/want to know. It can also generate images and in some cases videos as well.

[RabbidEwok]

It’s just ML rebadged because ML has gotten old. Real AI doesn’t really need to be fed data it can infer from unknowns like people can. You still need to pump tons of data into these “AI” to get anything from them. Also none of them have passed the Turing test. 

[jgreyz]

I hate AI. Thank you for this thread.

[Caroline]

Eh, it's not actual AI, only a bunch of learning models. There's more to intelligence than browsing through indexed data, these models are good at only one thing, say creating images, which can be fun, but GANs aren't actually intelligence of any kind, only intrincate software, a bunch of -if- lines is an oversimplification but you could see it that way, a prompt is a command that extrapolates data out of the index and forces the program to work only with the data that matches that string of text you've entered in order to generate something out of it.

Text models do something similar, translate simple commands like make this sentence into a formal email into something the program understands and print the data in screen for the user to copy.

 

I exclusively work with locally installed models that use the processing power my computer has to offer, and nothing else, for example this was originally an image of my room (source data) that was inpainted by the model using the prompt "decayed room"

Spoiler

it's not perfect as the colours have changed (the wood paneling is dark and the paint white) and the details like the radiator or light look weird but it's a great improvement from the green/purple artifacts I was getting before. I'm sure a human operator could also come up with something similar using the same source image and edition tools, but that'd cost money and time, and I can do that at home "for free", which is nice.

 

What I'm sick of is corporations calling everything AI, AI powered, AI backed, we're using AI for this and for that, etc. now that's tiresome and a bit lame because none of those things they're supposedly using are true AI, plus it looks stupid at this point, tho it's pretty much what brands do, remember when everything was blockchain powered? I do, where are all those things now? well, nowhere, because they didn't exist to begin with, the same happens with AI. Big companies are at most using discriminative models for nefarious purpose like processing hidden employee/customer credit and optimising quotas internally, I do know health insurance companies are 100% doing stuff like that to siphon more money out of people they deem prone to get sick due to comorbilities (obesity, heart conditions, mental health conditions, etc.), and the proper models can handle and process that amount of data MUCH faster than a bunch of office joes.

 

I'll limit my use to local models, will never use "cloud" stuff or programs that for some reason require an internet connection to work, an example of those are the hundreds of (mostly Chinese) shitware apps one can find in the Android app store, I don't use an smartphone but I know that happens, normies install all kinds of crap to their phones, who knows what those apps do not only behind the curtains but with all the private pictures people upload to their servers, it's all too shady... when something's free and seems too good to be true, distrust it, at least I'd distrust it... well. That's how it is. Another wall of text not made with Chat GPT.

[sub68]

I guess it preconceived notion

Depends, consumer yes its crap. But to me someone that has relatives doing projects with it for research or as a job its really interesting.

My dad did a research project in the early 90's basically was modeling human decision making, it got very close. interesting enough people still reference my dad's work.

On the other hand my uncle is kinda a cool he does machine learning mostly doing JPL rovers (almost every part) but the last 5 years he was doing cancer detection basically you take a sample and based on other samples it could detect what cancer it was or if there was any.

 

So there is examples of two different ai you have purely generative and analyzing.

For large analyzing of data bases it awesome and what really it should be for.