I've played around with a bunch of different variants of the AI hype. Gotta say, most of it is complete and utter crap. Sure, some of it is funny, and I can kinda see how the economy would shift, but the only reason assholes are panicking over it, is because it removes the useless parasitic jobs no one cared about.
Overblown? The idea that AI could write programs was a pipe dream for the far future. It's here now. And yeah, you're going to want a human programmer to drive it, but you're going to have AI doing the work of dozens of programmers. If you thought competing with Indian and Chinese programmers was bad, well good luck. You're now on your way to becoming part of the "useless eater" class.
The scary part is when they're going to stick it on a robot and give it a machine gun to kill the enemy with. You know this is going to happen sooner rather than later.
[ - ] chrimony 1 point 1.7 yearsAug 10, 2023 18:31:49 ago (+1/-0)
You need a person and always will, because you cannot trust the result.
So you have one person monitoring the results of AI doing the work of dozens of programmers, as I said. That's still an incredible leap forward from AI doing the work of ZERO programmers.
Yes it's overhyped. It gets a lot wrong.
Human programmers get a lot wrong too. But the AI gets more right on the scale of months, not years.
Also if you let it do to much for you - you become weaked an unhealthy
Of course, and this is one of the predicted dangers of AI. The better it becomes, the more it will be trusted, and the more it will takeover.
AI is only "overhyped" or "overblown" if you think it's ready now to go unsupervised without obvious faults. But the level of sophistication and capabilities being demonstrated NOW is incredible. We are on the cusp of a future that previously only existed in science fiction.
The scary part is when they're going to stick it on a robot and give it a machine gun to kill the enemy with. You know this is going to happen sooner rather than later.
Dude, you can do that now with off-the-shelf hardware. Basic digital actuator with a zip tie to the trigger of a gun mounted on a cnc'd platform attached to any basic tank-tread drone. Fuck it wouldn't be hard to build the drone itself if you know anything about fm/am circuitry. Firecontrol is a also all off-the-shell type hardware, or things you can pick up from a junkyard or off the curb. Lot of "ancient" hardware with perfectly good GPUs collecting rainwater by the side of the road these days. Motion detection and object identification can be done semireliably through opencv, and distinguishing friend from foe might be as simple as image-to-text applied to a bounding box (for each detected person) followed by feeding that into a offline-first RLHF-tuned LLM with prompts designed to classify each as "friend or foe" based on detected text features "wearing fbi uniforms and body armor", "looks like a politician or armed irs agent", etc.
System might be finnicky but it is 100% doable with existing hardware, and someone competent could build it for under $1000. Someone that already has the parts, experience, and has tackled other complex projects might be able to get away with building it for as little as $300-$500.
It'd be fun to build something like that not even to harm anyone, but just to be able to say you did.
[ - ] yesiknow 1 point 1.7 yearsAug 10, 2023 17:22:45 ago (+1/-0)
What it is, is a way to get people to stop using the internet. It's deliberately bad and turning people off. AI is just the jews flooding the entire internet with low quality AI crap. They don't want people on the internet getting real life information from real people.
IDK man. I watch everything Boston Dynamics do and those robots are pretty fucking intelligent. I know one day they will send them after us. A.I. is dangerous and once it advances so far, which I suspect the shit they have is, it becomes sentient.
Governments around the world already have vastly larger AGI programs than what the public sees. Looking glass was disclosed years ago as was the Russian equivalent. Likely a combination of organoid and quantum computing at this point.
[ + ] Bottled_Tears
[ - ] Bottled_Tears 5 points 1.7 yearsAug 10, 2023 14:49:10 ago (+5/-0)
I'm not worried because for the intelligent, it will be a tool to further get us ahead of the idiots.
[ + ] chrimony
[ - ] chrimony 4 points 1.7 yearsAug 10, 2023 15:06:41 ago (+4/-0)
The scary part is when they're going to stick it on a robot and give it a machine gun to kill the enemy with. You know this is going to happen sooner rather than later.
[ + ] Deleted
[ - ] deleted 0 points 1.7 yearsAug 10, 2023 16:25:05 ago (+0/-0)*
[ + ] chrimony
[ - ] chrimony 1 point 1.7 yearsAug 10, 2023 18:31:49 ago (+1/-0)
So you have one person monitoring the results of AI doing the work of dozens of programmers, as I said. That's still an incredible leap forward from AI doing the work of ZERO programmers.
Human programmers get a lot wrong too. But the AI gets more right on the scale of months, not years.
Of course, and this is one of the predicted dangers of AI. The better it becomes, the more it will be trusted, and the more it will takeover.
AI is only "overhyped" or "overblown" if you think it's ready now to go unsupervised without obvious faults. But the level of sophistication and capabilities being demonstrated NOW is incredible. We are on the cusp of a future that previously only existed in science fiction.
[ + ] prototype
[ - ] prototype 0 points 1.7 yearsAug 10, 2023 17:26:35 ago (+0/-0)
Dude, you can do that now with off-the-shelf hardware. Basic digital actuator with a zip tie to the trigger of a gun mounted on a cnc'd platform attached to any basic tank-tread drone. Fuck it wouldn't be hard to build the drone itself if you know anything about fm/am circuitry. Firecontrol is a also all off-the-shell type hardware, or things you can pick up from a junkyard or off the curb. Lot of "ancient" hardware with perfectly good GPUs collecting rainwater by the side of the road these days. Motion detection and object identification can be done semireliably through opencv, and distinguishing friend from foe might be as simple as image-to-text applied to a bounding box (for each detected person) followed by feeding that into a offline-first RLHF-tuned LLM with prompts designed to classify each as "friend or foe" based on detected text features "wearing fbi uniforms and body armor", "looks like a politician or armed irs agent", etc.
System might be finnicky but it is 100% doable with existing hardware, and someone competent could build it for under $1000. Someone that already has the parts, experience, and has tackled other complex projects might be able to get away with building it for as little as $300-$500.
It'd be fun to build something like that not even to harm anyone, but just to be able to say you did.
[ + ] yesiknow
[ - ] yesiknow 1 point 1.7 yearsAug 10, 2023 17:22:45 ago (+1/-0)
[ + ] Trope
[ - ] Trope 1 point 1.7 yearsAug 10, 2023 16:07:54 ago (+1/-0)
The most important thing AI can do is sort and filter data.
[ + ] HonkyMcNiggerSpic
[ - ] HonkyMcNiggerSpic 0 points 1.7 yearsAug 10, 2023 23:55:48 ago (+0/-0)
[ + ] Kozel
[ - ] Kozel 0 points 1.7 yearsAug 10, 2023 22:56:06 ago (+0/-0)
I've fired a copywriter I used to contract. Now I give chatgpt a prompt with a set of modifiers and it gives me for free something I used to pay for.
[ + ] Spaceman84
[ - ] Spaceman84 0 points 1.7 yearsAug 10, 2023 18:46:26 ago (+0/-0)
[ + ] Spaceman84
[ - ] Spaceman84 0 points 1.7 yearsAug 10, 2023 18:41:19 ago (+0/-0)
[ + ] AryanPrime
[ - ] AryanPrime 0 points 1.7 yearsAug 10, 2023 15:32:54 ago (+0/-0)
kikes glancing around nervously
[ + ] Doglegwarrior
[ - ] Doglegwarrior -2 points 1.7 yearsAug 10, 2023 15:33:33 ago (+0/-2)
Imagine 10 years after the car was invented and someone told you it would change everything?
Same with early computers.
Now imagine what ai can do in 15 years?
[ + ] albatrosv15
[ - ] albatrosv15 2 points 1.7 yearsAug 10, 2023 16:43:38 ago (+2/-0)
You are assuming progress, cause it's a white thing to do.