Artificial Intelligence (AI) thread

UFO

Chicken
I work with AI a little, not in my job but in my uni. I'm currently working on a game and I have no idea where people got this notion that AI is some kind of new 'intelligence'. It's just software, lines of code, it's not real, as in, it doesn't think for itself. It isn't a robot or a droid like on Star Wars. It can mimic human life just as a video game can portray drawings of people but it doesn't think or feel, it doesn't have motivations, it's essentially just a type of sorting algorithm.

I see the future applications for AI, we will be able to help the blind see, people who have lost their limbs will get better prosthetics. Medical technology improvements, space tech, virus elimination for Nano bots. There are great things that will come out of these types of programs but we're not talking any Terminator level stuff here.

I did a study recently on cybernetic prosthetics and we are getting closer to a Deus Ex future which is wonderful for people who have lost their limbs or who were born without. The world is much more accessible to someone who has a arm with fingers which can grasp things or legs that can bend and adjust themselves to the ground. This will mean no more wheelchairs or walkers, eventually people will be able to replace their bones so osteoporosis will be long gone as will Parkinson's. I any of you have ever worked as PSWs, you'll know just how hard it is for the elderly to get in and out of bed, it takes nearly 30 minutes.

Robots will create more jobs in the future than get rid of stuff. You know the ones that spout off about the AI eliminating work are the local commies. They love to live in this fantasy land where all work has been abolished because of robots and they are free to concentrate on their painting and drawing and all manner of flitty fairy things. The reality is that AI will bring us significant advances in machinery and tech. The forklift will graduate to the Power Loader, space travel will become a reality, farming and construction will be made better and more accessible. That said, we will still require people to operate these things, to colonize these worlds to oversee all the jobs. It's not going to magically erase all jobs. The internet created a ton of new work that we never new existed, AI will do the same... so long as we stay capitalist but that's a debate for another thread.
 

Oberrheiner

Pelican
Yeah you don't get it.

Do you understand how you train a neural network for instance ?
You basically teach it to reproduce what it saw.

So now let's say the government decides which student goes to which school programmatically.
This is already the case in several countries, you know ?
And of course the algorithms have been a major problem, badly written, difficult to audit and to let audit, controversial, etc.

Now just replace that by an AI, and poof, all your problems are gone.
The elite ensures it reproduces the established social order forever, and as the icing on the cake there is nothing to audit anymore.

You think the future is space travel, when it's a prison planet.
 
I work with AI a little, not in my job but in my uni. I'm currently working on a game and I have no idea where people got this notion that AI is some kind of new 'intelligence'. It's just software, lines of code, it's not real, as in, it doesn't think for itself. It isn't a robot or a droid like on Star Wars. It can mimic human life just as a video game can portray drawings of people but it doesn't think or feel, it doesn't have motivations, it's essentially just a type of sorting algorithm.

I see the future applications for AI, we will be able to help the blind see, people who have lost their limbs will get better prosthetics. Medical technology improvements, space tech, virus elimination for Nano bots. There are great things that will come out of these types of programs but we're not talking any Terminator level stuff here.

I did a study recently on cybernetic prosthetics and we are getting closer to a Deus Ex future which is wonderful for people who have lost their limbs or who were born without. The world is much more accessible to someone who has a arm with fingers which can grasp things or legs that can bend and adjust themselves to the ground. This will mean no more wheelchairs or walkers, eventually people will be able to replace their bones so osteoporosis will be long gone as will Parkinson's. I any of you have ever worked as PSWs, you'll know just how hard it is for the elderly to get in and out of bed, it takes nearly 30 minutes.

Robots will create more jobs in the future than get rid of stuff. You know the ones that spout off about the AI eliminating work are the local commies. They love to live in this fantasy land where all work has been abolished because of robots and they are free to concentrate on their painting and drawing and all manner of flitty fairy things. The reality is that AI will bring us significant advances in machinery and tech. The forklift will graduate to the Power Loader, space travel will become a reality, farming and construction will be made better and more accessible. That said, we will still require people to operate these things, to colonize these worlds to oversee all the jobs. It's not going to magically erase all jobs. The internet created a ton of new work that we never new existed, AI will do the same... so long as we stay capitalist but that's a debate for another thread.
You understimate the darkness of this world. The tech you speak is legale tech, not dark tech. Let's use the Resident Evil games like an exemple, in these games the Corporation named Umbrella use humans ginea pigs to advence its researchs, you think the powers of the real world will have scruple to use human ginea pigs and put a chip in their brains to totally understand how human intelect work and create an artificial version of it they can control even if they have to kill more than 1000 human ginea pigs to do so? They will not esitate. And when they'll succed they'll use corporation like Telsa, Amazon, Microsoft, Facebook etc to unlesh it to society.
 

Elipe

Kingfisher
totally understand how human intelect work and create an artificial version of it they can control
That's not how AI works, or at least, sufficiently advanced AI. That kind of thing would involve some machine learning models, and the thing about machine learning AIs is that they tend to be black boxes. It's very difficult to analyze exactly what a ML model "knows" because a ML model isn't based on principles, it's just based on a mathematical model of chains of inputs/outputs that align with a desired goal. To use an example, consider a ML model that learned how to play a Mario video game. It doesn't know what Mario is, it doesn't know what "enemies" are, it doesn't know what jumping means, or any of that stuff. It's literally nothing that you can really analyze in terms of learning principles from. It just knows that when it simulates a button press or a joystick push when given certain stimulating inputs, it gets better results.

AI doesn't work in terms of principles, like "dying in a Mario game is bad", it just works in terms of "oh, pressing that button at this time caused me to fail the simulation, that has been noted for next time". And then it repeats that process millions of times. That neural model gets massive, and teasing it apart to try to find the principles is a task that just won't be possible.

It's why AI actually sucks now, because when they turn the AIs loose, the AIs end up figuring out inconvenient truths from data that (((they))) really don't like getting out. Like for example, Amazon had that little trouble with its hiring AI figuring out that hiring men generally brought better work performance, so they just took it offline. Same with Microsoft and some of its AIs. I think Google had a similar mishap at one point too because it kept associating black people with gorillas.

They just take those AIs offline because there's really no way to make a good AI that operates on denying statistical reality. The problem there is that if they build an AI that, for example, prefers hiring women to men, it really does no better than just having a human HR director do the same thing. If anything, it actually performs (((their))) goal worse than a human would, because the AI wouldn't be subject to things like bribery or unprincipled exceptions.

They can't just "fix" those AIs, because there's nothing to fix. They whine that AI is "unconsciously" racist and sexist, but that's just how the real world data plays out. And they can't just go rummaging around in the ML models because again, it's all numbers that don't make any sense to humans. And if they try to build some rigged models using fake data, the problem there is that they're using fake data, and that renders the AI next to useless. And even if that worked out, they're going to want an AI that also adapts over time when it's released into the wild, but again, the problem is that the wild isn't fake.
 
What about quantum computing, considering all the recent progress within that field? If this technology is as powerful as I think it may be within twenty or even ten years, I suspect we will finally actually see Artificial General Intelligence developed.

What do you think?
 

stugatz

Pelican
This is sort of a side comment to this main topic, but has anyone seen those videos where an AI continues an already-written song? I actually can't listen to them without shivering (and I am not listening to this one a second time).


I don't know how someone can listen to this and not be completely creeped out. Mechanical, computer-generated type licks are one thing - like in half of what Daft Punk does - but this? I don't like listening to something a human didn't produce.

EDIT: Some people might note my Frank Zappa avatar. He definitely did produce some albums that veered into this territory (Jazz From Hell was near-impossible for humans to perform, so he fed the pieces into a machine). However, they still sounded like something, well...human. To an extent.

 

Stats

Woodpecker
Once you get good at it, you can make a killer living, possibly even remotely. And there's plenty of jobs out there in these fields, so a man with integrity can avoid working for ((Big Tech)) or the billy goat gruff.
curious what job opportunities are in nlp besides marketing and becoming a guru
 

Valentine

Kingfisher
Gold Member
...

It's why AI actually sucks now, because when they turn the AIs loose, the AIs end up figuring out inconvenient truths from data that (((they))) really don't like getting out. Like for example, Amazon had that little trouble with its hiring AI figuring out that hiring men generally brought better work performance, so they just took it offline. Same with Microsoft and some of its AIs. I think Google had a similar mishap at one point too because it kept associating black people with gorillas.

They just take those AIs offline because there's really no way to make a good AI that operates on denying statistical reality. The problem there is that if they build an AI that, for example, prefers hiring women to men, it really does no better than just having a human HR director do the same thing. If anything, it actually performs (((their))) goal worse than a human would, because the AI wouldn't be subject to things like bribery or unprincipled exceptions.

They can't just "fix" those AIs, because there's nothing to fix. They whine that AI is "unconsciously" racist and sexist, but that's just how the real world data plays out. And they can't just go rummaging around in the ML models because again, it's all numbers that don't make any sense to humans. And if they try to build some rigged models using fake data, the problem there is that they're using fake data, and that renders the AI next to useless. And even if that worked out, they're going to want an AI that also adapts over time when it's released into the wild, but again, the problem is that the wild isn't fake.

Their common counterargument to that is "junk data in = junk data out". I expect similarly to Google's SEO algorithms they're going to have to weight social justice factors more strongly in order to "balance the existing bias in our prejudiced society" :vomit:

I think one of the major relevant issues with machine learning that we'll face as men is that it makes policing thoughtcriminals much easier. China-esque Social Credit Score, facial recognition surveillance, writing analysis, "smart" cities with hidden microphones and internet access points everywhere for triangulation, etc.
 
I think my old PC was AI-driven. At least that was a reason I got rid of him. That slow crap always got stuck, frozen, or randomly restarted when I was in hurry or had something important to work on.
 
Top