Thread has been deleted
Last comment
AI TAKEOVER
Happy | 
United States KinglyFish 
This thing is 100% smarter than the average person. Just wait 10-20 years and it'll be smarter than the smartest humans. youtube.com/watch?v=PqbB07n_uQ4 AI TAKEOVER 2040 INCOMING
2020-11-18 13:21
Topics are hidden when running Sport mode.
Just say "This statement is false" and they break ez
2020-11-18 13:23
2 replies
#4
 | 
Germany iBUYPOLAND
+1
2020-11-18 13:25
This AI will probably start messing with you back if you try that. It's smart dude....
2020-11-18 13:49
Scariest thing is when AI starts creating AI. This will start happening in the next years and humanity has no ethics or regulations regarding this subject. I do not trust these 60 year old politicians to even grasp the concept let alone manage it correctly.
2020-11-18 13:24
15 replies
Nothing bad about it.
2020-11-18 13:30
12 replies
I am not suicidal enough to play Ukrainian roulette with AI. A mean spirited AI could very well be a good old genocide bingo and the end of humanity.
2020-11-18 13:33
11 replies
Lmao, if we ever develop a fully functional AI, it's way of processing information (that what we call thinking) would be completely different from the human one. Since it doesn't have a innate instincts which would urge it to fight with other species to secure its physical survival. It's actually the most common layman's mistake - perceiving AI by human standards. It would be completely, but COMPLETELY different.
2020-11-18 13:38
10 replies
I never claimed to know how a potential AI would think. That is exactly what you are doing right now about a technology that is still in the very early stages. It is definetly an amazing technology that could benefit humanity greatly. But it is also very dangerous and needs to be treaded lightly. I just pointed out that there are currently no safeguards in play to prevent an AI disaster. You are the ignorant one in this matter.
2020-11-18 13:46
6 replies
Sorry, but if you don't believe old politicians that rule you in everything and taken that genocides did and do still happen fearing AI doesn't make a lot of sense.
2020-11-18 13:56
2 replies
Politicians wont genocide the whole human race at least. I am not fearing AI. I am just saying it needs safeguards. You are too ready to bow down to an AI overlord while expecting this will result in an utopian world with all your problems solved. This does not make sense either.
2020-11-18 14:00
1 reply
Dude, i deal with the Ukrainian officials, not with the Dutch ones. That should explain a lot LMAO
2020-11-18 14:02
It's only as dangerous as bad architecture allows it or a bad person wants it. What would be troubling is not the AI perse, but the autonomy given to it and even then, it would require a series of directives and patterns to be defined making the AI able to decide whether to chose from one or another path after working a specific set of information. AI can't build physichal things for example unless you merge it with autonomous robotics and still, they would limited by their energy and space to do any harm. Let's say that a concept like Skynet or HAL, where an AI would take over everything, in reality most likely it would only be able to kill sattelites and internet, things that could, in theory, be overrided back to safety fairly quickly. Even this, could be prevented by a "simple" directive on the AI code. Have you ever watched that movie A.I.? That 3 law concept is actually more accurate than AI grewing uncontrolably
2020-11-18 14:00
2 replies
Adding to it, i think the most harfmul way an AI could hit us is by convicing us that their information is right when they are not. This would make us take bad/wrong decisions and this is potentially troublesome.
2020-11-18 14:02
1 reply
Your comment about bad architecture is exactly why we do need safeguards for AI. But the biggest problem will come when AI develops a more advanced AI that is when things can spiral out of control fast and the oversight and transparancy could be lost. There is dozens of ways that an AI could cause massive problems by having acces to the internet. Miss information is definetly one that comes to mind. No, I have not watched any movies about AI at all.
2020-11-18 14:16
This may be true, but it depends on the training data and the architecture. You can't exactly control the conclusions an AI comes to, explained further here: #38. A overlord AI may come to the conclusion we're useless or it may just decide its purpose is to guide us.
2020-11-18 14:13
1 reply
Well, probably I'm not that fond of people in general lol
2020-11-18 14:24
AI is created by humans tho
2020-11-18 15:22
they govern your life already
2020-11-18 15:20
1 reply
True, and so far Governments have failed famously with limiting big tech's power.
2020-11-18 15:46
Terminator coming true...
2020-11-18 13:24
#5
shox | 
Switzerland Log@n
Incoming AI versus AI csgo ez No mistakes and perfect aim ez
2020-11-18 13:25
#6
RpK | 
France pol44r
finally.
2020-11-18 13:26
Stop watching terminator......
2020-11-18 13:26
9 replies
Did you watch the video? The AI tells jokes, can mess with people, tells lies 'when it's to its own benefit', etc...
2020-11-18 13:41
8 replies
I did. Stop calling this pre-made code answers AI.
2020-11-18 13:43
7 replies
That's not what this is AT ALL. You can research gpt-3 yourself. It's answering all of this live and can even its own code.
2020-11-18 13:45
6 replies
Jesus christ. You really don't understand anything about this.
2020-11-18 13:47
5 replies
Guy who has literally no clue what he's talking about tells me it's all "premade code". Projection much. Do you even know what a deep learning model is?
2020-11-18 13:48
4 replies
Yeah i have no clue. 0/8
2020-11-18 13:51
3 replies
Okay enlighten me on exactly what you mean. Inb4 "the architecture needed to be coded, therefore it's pre-coded and doesn't count".
2020-11-18 13:53
2 replies
Wtf are you even talking about.
2020-11-18 13:54
1 reply
Lmao. I thought you meant you actually you knew something about this with that "0/8". You've got to be baiting. How're you so confident while having absolutely 0 clue what you're talking about? I'm legitimately curious.
2020-11-18 13:57
Pee on them
2020-11-18 13:27
Would be nice having an AI governing a country instead of those baiting mofos
2020-11-18 13:29
3 replies
It'd be great until they decide humanity are "redundancies" to efficiency.
2020-11-18 13:43
2 replies
#23
 | 
Singapore Am2de
they are not wrong though
2020-11-18 13:50
Who said they'd be chasing efficiency?
2020-11-18 13:51
#11
 | 
United Kingdom DJ_Nr8
8:10 RIP HUmanity
2020-11-18 13:31
#15
 | 
Portugal xxxruixxx
Skynet is coming. Brace yourselves,
2020-11-18 13:42
An AI is just a bunch of if else, nothing more. The only reason an AI could kill someone is if the human who built it is a killer.
2020-11-18 13:51
2 replies
People who build AI's don't actually know exactly what's going on in them. They just build the architecture and let the AI train itself on the preselected training data. The scientist can curate the training data, but especially on more complex models they can't exactly control the conclusions the AI comes to from that training data.
2020-11-18 14:02
1 reply
Sure like i never know what the random fonction is gonna give me, but its certainly not random and i can easily influence it. Doesnt matter what the AI learns, it's still a compilation of if else behind everything. He learns what we allow him to learn.
2020-11-18 14:38
Why would they make a robot that needs glasses and have Weird Al hairstyle.
2020-11-18 13:52
Sorry man, but this will never happen (atleast not before the humans destroy themsenves) . They just have more info, but nothing is really espontaneous and they cant create trully new knowledges.
2020-11-18 13:53
5 replies
It's hard to tell right now since AI development is still in its early stages. The current weights and biases, layer/architecture based deep learning may only take us so far towards general intelligence. But, like all things another method will come along and change everything.
2020-11-18 13:56
4 replies
Imagine the world without Einstein and Hawking. I cant imagine any AI making discoveries like these guys did, just because they made it years before the proofs of the theories and thinked about things really different of what we knew about the universe.
2020-11-18 14:06
3 replies
The world is very business oriented, so 99% so AI's are researched and designed for business purposes. I think that's the only reason more people aren't trying to develop a true general intelligence. But, hypothetically if you built an AI with architecture that mimics the human brain and trained it to have a "general intelligence", then that AI can not only come up with unique ideas it can also program another AI.
2020-11-18 14:06
2 replies
Thats the point of "atleast not before the humans destroy themsenves" on my first statement. Maybe is possible to make a structure that mimic the humans brain, but i think it will be something biological and not computational. Maybe a AI needs the biological component to react apropriately to the environment, but it will be like creating life or cloning, so i think it's not what we are talking about.
2020-11-18 14:17
1 reply
Why does it need to be biological? With future computing power we may be able to simulate a full human brain in a computer program.
2020-11-18 14:19
When he says that lie when is interested on it, why he didn't lied? Kkkkkkkkkkkkk intelligence
2020-11-18 13:58
AI is just a buzz-word from tech-firms that want to sound advanced to regular people.
2020-11-18 14:01
1 reply
Yeah 99% of AI used are just stats based regression models or something simple like that. But, there are true "general" intelligence AI's being developed right now like this one. Did you watch the video btw, it may change your mind.
2020-11-18 14:08
Lol it's so funny people thinking "OMG AI is gonna take over"..if you know anything about AI, we are very very far from that day. Furthur, don't think u are the only one concerned regarding AI safety, there are people dedicating their career to it: doing AI research in a safe way...so just chill and stop being the person who talks without knowledge
2020-11-18 14:13
8 replies
Dude I work with AI, it's not like I'm clueless lmao. Pretty much anyone can train a competent chatbot these days, even if its a super basic bare bones RNN based one. What's the version of that in 20-30 years: anyone can release their own malicious 200+ IQ AI onto the internet or the world? Think a bit more broadly and see the possibilities.
2020-11-18 14:24
7 replies
Okay, well sorry if i was harsh..good to know u work in AI. Okay....yes...there is a lot of technological improvement if you consider what we have achieved in last 30 years ago...however that doesn't imply we are going to discover new things at the same rate. We are easily at the stage where we learn more(deeper neural networks) as we have more computation power, however computational power is getting plateaued. Big discoveries in Quantam Computing may change a lot of it, but I don't think we are anywhere near "General human Intelligence" and we won't be there anytime soon..at least till 2040(as u mentioned in the post)
2020-11-18 14:22
6 replies
I mentioned in some other posts here on the limits of the nodes/layers classical deep learning style of building AI architecture. At the end of the day, someone may rethink how AI is developed and it'll change everything. Or there's a huge overhaul in how we develop classical deep learning architecture and it changes everything. Right now it may take 1000's if not 10's of thousands of attempts to create a single excellent deep learning architecture. If that process is sped up massively, we could create enormously complex models that easily eclipse the human brain in terms of computational power and "general intelligence". Right now even the most complex AI models are built on relatively "simple" base architectures. Either way there could be a revolution and the AI rapidly overtakes humans.
2020-11-18 14:31
5 replies
I understand what you are saying...right now the best models are just neural networks that are just extremely deep(GPT-3 for example)..However, when you say "At the end of the day, someone may rethink how AI is developed and it'll change everything", I say that's a very big thing which is unlikely to happen just suddenly. Slow improvements are more probable than a sudden "HUGE" discovery...and slow improvement will take a long time...a HUGE discovery is possible..alghough extremely extremely unlikely, that's all I'm saying. And I've already said that if developing deeper models is the way to go, then we can't reach very far using that as well because of technological limitations
2020-11-18 14:34
4 replies
Yes, one part I left out of it is for "classical" deep learning to be able to grow more complex, the hardware will need more computational power. And on top of that the process for discovering new deep learning models will need to be accelerated somehow, right now the process is way too labor intensive even for relatively simple architectures. What I mean by a "huge" discovery, is someone uses a new method to create an AI that's traditionally coded using deep learning and this new method works better than the traditional deep learning. It doesn't have to be something "huge" or major, just a beginning. And it's really limitations in parallel computing that's holding back AI right now from going above traditional deep learning.
2020-11-18 14:40
3 replies
Okay, yes that kind of a "huge" discovery is totally possible. However I'd say the thing is, we are far from what we may call "General Intelligence". GPT-3 works good for language, but it's far from being something of general intelligence in the textual domain. Same goes for image domain. Same goes for control tasks, we have good reinforcement algorithms, but nothing general, a lot of tuning is required for these algorithms to work for simple control tasks. That's why I don't feel the need to worry about "Evil AI" just yet, also slow improvements on that huge discovery will also mean that researchers will keep the safety aspect in mind. However yes, it'll be interesting if a general intelligence method gets developed soon lol
2020-11-18 14:47
2 replies
Yes language based models are somewhat "simple" since language is inherently pattern based and there are many excellent architectures based off the classic old RNN. General intelligence is definitely far away, but I don't think it can be developed using current deep learning. It'll need to be something new and will require stronger hardware than modern GPU's (for training) and others can provide.
2020-11-18 14:52
1 reply
Yeah...I think reinforcement learning algorithms(using neural networks perhaps) will be the right way to go ahead in order to achieve "General Intelligence"(because I believe human intelligence is largely based on the reward/punishment system) So what do you do, are you a student? Or you have a full time job as a data scientist or AI engineer or something?
2020-11-18 14:56
#56
 | 
Canada JC_123
Simply turn the PC off EZ
2020-11-18 14:43
#60
 | 
China Manchu
It is scary and interesting at the same time, what will they do and when but also raises a question about certain jobs.
2020-11-18 15:05
prntscr.com/vlir5q yes cleverbot wil take over the world soon with these average 2 word answers
2020-11-18 15:08
1 reply
Man cleverbot is from the 90's. Even I can program something comparable to cleverbot in one evening.
2020-11-26 16:25
doesn't sound like some to be afraid of on the other hand that weird guy sounds dumb
2020-11-18 15:18
#65
 | 
Ireland tka4
AI bots can't even recognize pandas imgur.com/a/0Ks8ajF
2020-11-18 15:25
3 replies
Obviously if you apply some filter on the picture that's meant to trick it, it'll get it wrong.
2020-11-26 16:42
2 replies
#69
 | 
Ireland tka4
Not very superior then, if it can't adjust to such basic noise
2020-11-26 16:59
1 reply
Anyone can make a object detector in one evening with Tensorflow. There are better models out there that won't be fooled by a basic filtering layer or some noise.
2020-11-26 17:02
Scary. Damn
2020-11-26 17:04
Winstrike
1.47
Tricked
2.58
OG
1.52
Dignitas
2.62
Sangal
1.30
HONORIS
3.32
Bet value
Amount of money to be placed
Winning
Odds total ratio
-
Login or register to add your comment to the discussion.