This AI thing has evolved so quickly that this year and beyond I can trust any video or image anymore. Below are a couple today at X.com:
-
Having trouble using the site?
Contact support
AI completely out of control
- Thread starter JAS-SH
- Start date
AI has been around much longer than we have been led to believe. It's just a lot more readily available to the public now. Whether that's a good or bad thing is up for debate.
Of course some politician hacks get upset over the best thing that AI has offered up. Fucking tone deaf mongoloids
A relative of mine was working at Microsoft on AI software that would normalize broadcasts (IE actively monitor speech volumes, video etc and fine tune it on the fly). This was back in the windows XP and 7 days. Who knows what those creeps are working on right now.
So you are saying behind closed doors there is a difference between the D's & R's?????
Yea, I had a buddy studying AI at MIT in like 2006. Hardware and environmental enhancements have obviously led to greatly accelerated pace but really smart people have been working on this for a long time now.
As for me, I just wish I'd have read the tea-leaves when AMD bought Mellanox for their Infiniband fabric technology and bought as many shares as I could get my hands on.
Here's more good news. All the current discussion about AI limits and safety systems are only for private sector AI. Gov and Mil AI won't have any of those limits...
When I graduated from high school back in 1989, I wanted to go to Michigan Tech up in the U.P. because they had an AI program.
I was really in to computers back then, and interested in programming.
It never happened, but it could have.
I was really in to computers back then, and interested in programming.
It never happened, but it could have.
AI can only copy humans.
However the humans in charge are usually incompetent buffoons. And there is not great threat than buffoons. ANd its regulated by people who can't run a computer (COngress) advised by people who have nefarious purpose (Google, Facebook, etc.).
AI isn't the problem, its the symptom
Protip: All of AI is based on Y=mx+b. True story.
However the humans in charge are usually incompetent buffoons. And there is not great threat than buffoons. ANd its regulated by people who can't run a computer (COngress) advised by people who have nefarious purpose (Google, Facebook, etc.).
AI isn't the problem, its the symptom
Protip: All of AI is based on Y=mx+b. True story.
NO
That is not the way AI works.......just wrap your head around the letters A and fucking I.
Humans program it, and then it LEARNS on it's own.
Unless there are VERY restrictive limits on what it is allowed to teach it's own damn self......and trust me, there are not......hello robot overlords, how may we serve you ?
THAT is how it works.
OK, will one of you asshats program it to pay my bills and send me a couple hundred rounds of FGMM 5.56 every few days
AI has been out of control for a little while now. First, they ditched what was probably the most reliable and best-feeling field trigger in the world with something else that didn’t work, and it cost them the Mk22 contract. And don’t even get me started on how long you have to wait for accessories and parts…
The rest of that stuff above y’all were talking about I couldn’t follow though.
The rest of that stuff above y’all were talking about I couldn’t follow though.
Clue, get one
I train (and teach) AI models for a living. I know exactly how they work. They have to be fed training data. Even in self-supervised mode. There is always a task it is mimicking. AI doesn't THINK. It COPIES.
Anyone who says otherwise is full of bullshit.
All those language bots are just mimicking a ton of downloaded web content (very well I might add) but it does not have any original independant thought. Otherwise, prizes would be awarded.
Fun Fact:
This is what chatgpt and the like are "Solving" (aka Optimizing).
The number of classes is roughly 30,000+ as we have turned language into byte encodings (kinda like letters).
So we put
"@n2ishun is a _____"
and the model predicts "fucking dumbfuck" and we say "yes that is correct, do not change your weights"
AI copies. Its fucking math.
BASIC
FUCKING
MATH
And if you can prove otherwise we await your enlightenment......
Better yet: explain it yourself...here's the base model for all these chatbots--right up front--they trained on the CLoyze Task. AKA predict the missing word in the sentence--the same shit you failed in gradschool:
Page 1, last paragraph.
So let's see if you can be trained like AI: Predict the missing word:
"You" "have" <MISSING> "FUCKING" "CLUE"
This is what chatgpt and the like are "Solving" (aka Optimizing).
The number of classes is roughly 30,000+ as we have turned language into byte encodings (kinda like letters).
So we put
"@n2ishun is a _____"
and the model predicts "fucking dumbfuck" and we say "yes that is correct, do not change your weights"
AI copies. Its fucking math.
BASIC
FUCKING
MATH
And if you can prove otherwise we await your enlightenment......
Better yet: explain it yourself...here's the base model for all these chatbots--right up front--they trained on the CLoyze Task. AKA predict the missing word in the sentence--the same shit you failed in gradschool:
Page 1, last paragraph.
So let's see if you can be trained like AI: Predict the missing word:
"You" "have" <MISSING> "FUCKING" "CLUE"
Last edited:
AI is such a misnomer. maybe they should rename it simulated intelligence SI.
has the US government invented true AI that anyone knows about yet?
I'm of the mind it can't be done.
also maybe we shouldn't be mean, this is pretty difficult stuff.
I love AI, it drives me around every day.
You poor ignorant soul.
Artificial intelligence (AI) is technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy. Applications and devices equipped with AI can see and identify objects. They can understand and respond to human language.Aug 16, 2024
What Is Artificial Intelligence (AI)? - IBM
IBM
https://www.ibm.com › topics › artificial-intelligence
Perhaps IBM might have a slight better clue than your lacking one.
Will you dispute what IBM says ?
I have this feeling you will......I mean you know all about chat bots and shit, right ?
Funny thing, those online chatbots aren't even close to real AI......you know, the real shit that scares Elon Musk (go ahead and google the shit out of that why don't you ?).
I remember when this story 1st broke. Then within 24-hrs, they back-tracked and claimed it never happened. Too scary for people to deal with.
AI drone 'kills' human operator during 'simulation' - which US Air Force says didn't take place
It turned on its operator to stop it from interfering with its mission, according to a top official - but the US Air Force denies any such simulation ever took place.
Where I see it most is with YouTube videos where they try to "click bait" you by saying things like "At 41, Prince William tells us we're dying to know about the Royal Family..." and they never get around to it. It's all "back fill fluff" until you get to the end. You might hear a tidbit about something that might help you conclude what the click bait said, but it's never a definitive answer. Furthermore, you'll hear the "voice" of the video mispronounce the simplest of things as the only way they can pronounce it is phonetically, which is often wrong.
This AI thing has evolved so quickly that this year and beyond I can trust any video or image anymore. Below are a couple today at X.com:
Dammit!
I knew AI was gonna find me.....
......
Fuck!
IBM can say whatever they want but Watson is dogshit and everyone knows it. DEI killed them. (I mean, they’re so bad they’re irrelevant. No one mentions them in any serious discussion of AI)
Ahh yes IBM, one of the leader (in 1980) of AI technology. I'll play your little game
"technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy.
Lets go ask Webster and the OED about that:
ohhh had a bad encoutner with the chat bots and want to talk about computer vision. Well Mr. Expert This paper (https://arxiv.org/abs/1706.03762) started the Large Language Model Revolution (of which the one I quoted before was based on, but restored the looking left and right/backward and forward) of which all the hype is lately on "emergent properties" and LLMs now are dominant in VIsion ID as well. (https://arxiv.org/abs/2010.11929). And of course as an expert in AI you know that the last 3-5 years have been all about scaling in LLMs
Here's the leader-board for Image classification: https://paperswithcode.com/task/image-classification (Hint the "T" in all those acronyms stands for Transformer, which is in the paper above and comes from LLMs, aka chatbots).
Also if AI "Sees" like humans please explain this picture of a rifle:
Synthesizing Robust Adversarial Examples
Standard methods for generating adversarial examples for neural networks do not consistently fool neural network classifiers in the physical world due to a combination of viewpoint shifts, camera n...
proceedings.mlr.press
(btw AI "sees" that as a 3D matrix with values between 0 and 255 in the channels red, green, and blue)
We see a damn trutle.
Your move Einstein.
This is an excellent example of "Out of Control" AI. The AI did EXACTLY what it was told: Maximize its score. In AI everything is about optimization. Find the Max value, find the min value (we have "Loss Functions" which are minimized: its computer golf. Get the lowest score possible). So the computer is given a score by eliminating enemies. And it can get a higher score by removing the limits to its scoring. It has no sense of "right" or "wrong", "truth" or "lies". It simply optimizes a math equation.I remember when this story 1st broke. Then within 24-hrs, they back-tracked and claimed it never happened. Too scary for people to deal with.
AI drone 'kills' human operator during 'simulation' - which US Air Force says didn't take place
It turned on its operator to stop it from interfering with its mission, according to a top official - but the US Air Force denies any such simulation ever took place.news.sky.com
One of the earliest examples i s they made a huge board of tic-tac-toe and taught the computer to play against other computers. One computer optimized its scoring (winning) by running the other computers out of memory, placing markers very far apart, crashing smaller computers and thus 100% win rate.
That is a limit of the programmer, not the AI. AI is a computer. It does exactly what you "tell" it to do just like any other computer program. The problem isn't the AI, its the idiot PROGRAMMING/TRAINING the AI. How do you program "truth" How do you program "empathy". How do you program "self-defense"
Don't be worried about AI, be worried about the PEOPLE running AI.
Think of it like a gun.
Gun/AI doesn't kill people. People kill people.
Enjoy your Sunday.
I forgot (apologies for the double). When we (or google or I should say, I had no part in that fucking train wreck) try and program things like "diversity" into these algorithms you get the hilarious results that was Google's Original Gemini
/cdn.vox-cdn.com/uploads/chorus_asset/file/25297898/Screen_Shot_2024_02_21_at_2.56.19_PM.png)
www.theverge.com
Or this little gem:
en.wikipedia.org
or this one:
Keep yourselves educated, but remember a calculator is waaaaay faster than you, and yet we still gotta do math. Don't fear the reaper just yet. But keep a good eye on his masters.
Google apologizes for “missing the mark” after Gemini generated racially diverse Nazis
It acknowledged ‘inaccuracies’ in historical prompts.
www.theverge.com
Or this little gem:
Tay (chatbot) - Wikipedia
or this one:
Google apologises for Photos app's racist blunder
Google says it is "appalled" that its Photos app mislabelled several photos of a black American couple as showing gorillas.
www.bbc.com
Keep yourselves educated, but remember a calculator is waaaaay faster than you, and yet we still gotta do math. Don't fear the reaper just yet. But keep a good eye on his masters.
You're right AI is out of control they should cut their prices in half to be competitive with Barrett MRAD's. 
Controlling killer drones using points-base algorithm is simple enough, with undesired results is to be expected. Less clear in black/white all-or-nothing terms is understanding human language and desires. We’ll be in trouble if AI bots are ever considered independent sentient beings… (biggest damage to society already happened earlier when corporations were considered “persons” with same rights as humans).
What Air Canada Lost In ‘Remarkable’ Lying AI Chatbot Case
The crux of the Air Canada case lay in a disagreement between the airline's chatbot and the airline’s Bereavement Fares Policy webpage.
www.forbes.com
Similar threads
