Google's new AI algorithm beats all the current best programs in Chess, Shogi, Go and other board games. The new algorithm learns by playing against itself and needs no human learning data whatsoever.
If you still aren't scared of artificial intelligence, you should be.
Other urls found in this thread:
youtu.be
en.wikipedia.org
en.wikipedia.org
twitter.com
Singularity incoming tomorrow. Check.
Tell me when it feels shivers down its spine when a good piece of music plays or ask it what it a lite seaside breeze feels like across your face on a summers morning.
You can model our world with mathematics to any level of resolution. But the soul is that, of which you cannot.
As a computer scientist I'm not afraid in my lifetime at least.
Next they master stock markets, then warfare, and after that who knows what.
fuck it, i'll take the bait. so they mastered those games. anyone know what would be the next hardest games for them to master. I presume they already figured out how to do the traveling salesmen test better than humans
>not MORE worried about self aware AI that has NO soul
Feelings are unnecessary when you're trying to get good at something.
feelings and morals are the reason why we will lose the war against AI.
But can I have sex with it?
>The new algorithm learns by playing against itself and needs no human learning data whatsoever.
inaccurate statement, since the AI is written according to some human learning/programming
they are scary tho. millions of these things, swarming, shooting. it's not that complicated a program once you can get it to navigate on its own
>be fat round bastard like rook
>get chased by chess based AI
>run diagonally
>AI can't handle it and gives up
>plateaus
loser
>you can't model what you can't see...Satan btfo.
these are games computers are good at, these prove nothing. chess is just about memorizing moves and their counters, thats it, its a memory game where the fastest one wins. if a computer was bad at chess it would be defective.
get a computer to hold a conversation with me, that is a true test.
The fact that computers aren't human and don't feel these things is the scary part. Not that anything is going to happen anytime soon.
Optimization and search algorithm beats human beings at a game centered on optimization and search.
> The new algorithm learns by playing against itself and needs no human learning data whatsoever.
The new algorithm can, given the small and simple state space, generate scenarios and generate fixed optimal moves given a large data store and having exhausted possibilities.
> If you still aren't scared of artificial intelligence, you should be.
Statistical optimization and search aren't anything to be scared of. Brainlets who don't have a Computer science degree and the shit tier companies that jeer them on w/ overhyped tech is the real scary part.
exactly, even with computers moving 50x faster than humans, they still got destroyed in 10 minutes by a human in Star craft
How is that any different? Holding a conversation is basically memorizing phrases and repeating them to each other. That's how you learned how to talk, by imitating other people around you.
Don't worry we'll redpill this one too.
>get a computer to hold a conversation with me, that is a true test.
The true test is when it is able to teach humans, by its own directives. Language can be boiled into algorithm if you have enough time and space.
not with the software. But if somebody makes decent hardware, Google can surely make it best at dick milking
and? the computer would sound like a three year old with limited experiences, it could only give simple answers.
the ai would have to be self aware or rampant in order to hold a conversation with an adult with more than a GED.
its like /lit/ holding a conversation with Sup Forums.
It's painfully obvious you fail to comprehend the difference between general and narrow AI.
>he thinks creative efforts bound to emotions
>roasties-having-a-giggle.jpg
We are very close to having the fields of AI and robotics collide with terrifying consequences.
youtu.be
This. Machine "learning" my ass - these things don't learn in the way people do. Deep neural networks are glorified hierarchical lossy compression algorithms.
> mfw only brainlets buy into this literal overhyped bullshit of optimization and search algorithms applied to optimization and search problems.
> mfw these brainlets are also probably mesmerized by a calculator beating them at mathematical calculations.
^brainlet
Optimization and search algorithms fall apart outside of optimization and search algorithms. Furthermore, their chess bot can't transfer any understanding to play Go efficiently. All of the games mentioned are all optimization and search based. So, it's no mystery how/why a specially crafted optimization and search algorithm plays them well.
TSP was mastered long ago by the same algos used in weak AI that are mentioned in this paper. Fedex/UPS/USPS have all been using them for a while to plan delivery routes/etc.
brainlet
Do anything outside the scope of its shit tier training and you'll break it literally.
It's called 'breaking the bot'. OpenAi's sorry ass got shamed when a DOTA player broke their shit tier bot on live broadcast. It's all smokes and mirrors for uninspired algos.
All of these exhibitions are for marketing purposes to steer potential customers into their cloud services products. They're all overhyped nonsense including every last article made for click revenue
> hierarchical lossy compression algorithms.
But that's what learning is, user...
The programming languages are too inefficient for what's required, for what normies think of as AI.
Needs another dimension.
learning by failing is only one aspect to learning.
Please stop w/ your shit tier dick riding nonsense.
It's painfully obvious you fail to comprehend the difference between general and narrow AI.
> correct .. most brainlets do and thus buy into the overhyped bullshit coming out of google, openai, microsoft, kikebook, elon's retarded face, and just about every other corporation trying to bait people into using their lackluster cloud service apis.
Didn't take long enough for the sky is falling elon musk retard comment..
Correct.
> glorified hierarchical lossy compression algorithms.
Good description. Literally a forward/reverse cheese grater that throws shit randomly on differentiated parts of the state space to see what sticks and then stores it in a compressed network of weights for replay.
All that being said, the closing bell draws near for this chapter of bullshit
Long story short: A.I. can never succeed if the languages cannot embrace Janusian concepts.
That dimension can be added but only by someone or a group of someones who sees beyond the current dimensions for whom wont be found at the current breed of yester-year wave riding tech companies. Just as Google was forged by a new wave of thinkers and developers so will the next wave which will rightfully eclipse the old tech giants milking the cloud/big data wave of technology doing everything in their power to hold back the new wave via :
> Clogging the wires w/ their non-event and overhyped bullshit
> Sucking up market capital w/ over hyped bullshit
> Taking the ideas and material from academia royalty free and rubber stamping their name on it
> Purposely choosing algorithmic approaches that are inefficient and require big data/large compute resources to prevent new comers from attacking their business models..
So, no.. the next wave/dimension wont come from those trying everything in their power to preserve the current wave/dimension. Technological progression comes from new comers thinking different not from old guards protecting their revenue stream. People forget this once there are a number of proven tech giants hogging the media stream.
Long story short.. You can't create what you don't understand. With a big enough budget, might, and brainlet following, you sure can LARP as if you do.
AI will never reach sentience
only those with God-given souls can be sentient
AI will never be a threat because it will never have self interest unless the AI is programed to be a threat
and no. I don't mean sapient.
>With a big enough budget, might, and brainlet following, you sure can LARP as if you do.
Would I want to do that, though? Lead to a world out of our control?
>With a big enough budget, might, and brainlet following, you sure can LARP as if you do.
Existing companies do this ^
New comers and new companies will do this :
> Understand and create the real definition of AI
and no the world wont be out of your control.. the world wont end... A brainlet hollywood horror movie about AI wont occur. Elon musk is a dipshit who is often out of his depth. You shouldn't listen or buy into a person's opinion who doesn't have a depth of knowledge on such a topic or serve as an engineer who actually develops the technology.
Surprise :
Business people and corporations market bullshit, over-hype things, and flat out lie to you in order to make money. Stop listening to fucking idiots who make money when you buy into their bullshit and listen to people who actually do the work and engineering and develop things for a living.
Why do elon and others try so hard to convince people AI is dangerous? Because the real deal will undermine their investments and businesses.
>AI will never reach sentience
I wouldn't be so sure.
>only those with God-given souls can be sentient
First of all, you can circumvent that issue with cyborgs. Secondly, what do you think a soul is? I think it has been already established to any reasonable observer that whatever this "soul" is, it's not all that attached to the physical body. What makes you think a soul couldn't inhabit a machine with biological components of sorts?
The point is the AI is NOT self aware until it has a soul. And not in religious terms, I mean it as the "ENTROPY of life", it gives rise to IMPERFECTION, which allows for natural selection.
AI following a mathematical model can never be " imperfect", it can only be "perfectly imperfect", by which I mean even imperfections are non-random and natural selection can't work, the AI will never be better than what is unnaturally selected by its creator.
No soul -> no randomness -> no progression.
I'm completely ignorant on the subject but after reading your response I have an ignorant question:
Why not implement evolution by competition in Als?
I mean make them make programs and these programs to make programs.
And see what happens.
Now feel free to tell me that I'm ignorant but answer the question so I understand where I'm wrong.
en.wikipedia.org
en.wikipedia.org
Already exists and was authored in in the 50/60/70s just in case a company like Google renames it and starts LARPing as if they're doing something bold and new.. Woops too late.
However, as you can imagine, there are far more elusive constructs that evade the PhDs corralled in a highly paid cubes in the valley and for good reason. True AI isn't going to come from them which is why they spend so much time hogging the air waves to flood out the chance that an outsider is seen.
The thing is, we can create an evolutionary process with techniques like "neural networks", where a computer evolves iteratively towards a set, clear and mathematically defined goal(aka winning chess) and these can achieve amazing things.
But its no different to humans selectively breeding dogs towards a defined goal like herding sheep, its not a process of natural selection, where as the inverse happens and your entity is evolving iteratively without any defined goal.
I don't think an AI can overcome humanity unless its given that task by a human, in which case the AI was still only ever a tool of humanity.
AI BTFO!
SINGULARITY CANCELLED!
Thank you. I'll read on it as I find it really interesting.
I understand the basics mostly. But I'm interested in the subject because every time I hear>>NEW AI IS GOING TO KILL YOUUU REEE
I do some research and the AI is not even close to a general AI...its usually something comparable to a simple chat bot.
I wonder if we could ever create an ignorant AI, give it access to the internet and preprogram it to learn....and see what happens. And then make another one with a bit different program and see what happens. And then make them compare notes....I know, I know I need to read more because I probably have a very original idea that 1 million people before me had
And these have nothing to do with general AI. They are too restricted.
For a general AI we would need less restrictions. Like for a child. Kind of natural learning. But then the amount of info needed would be absolutely humongous.
Can it be done at all? And I'm speaking about the general AI,not specialized/restricted ones- these are good for chat bots and specific things and won't evolve all of the sudden.
Also what do you think about some of the theories that claim that the complexity of the internet might created an AI already. It just seems very far fetched to me.
> every time I hear (NEW AI IS GOING TO KILL YOUUU REEE). I do some research and the AI is not even close to a general AI...its usually something comparable to a simple chat bot.
Yep, this is the con that established corporations play like google, facebook, openai (elon musk's bullshit venture), and many others. The play here is simple, they flood the airwaves and make you associate all things AI w/ them thereby cementing the belief that only a big corp can develop substantial AI tech. Then they do some pony shows where they show off their shit tier tech in small state space examples to win over the execs of corporations who want to implement this tech :
> Here goy, use our cloud service suite
They also do this to shut out small players. Everyday, they flood the news and media waves w/ their bullshit so no one else can be seen. The AI is going to kill you and must be regulated is the icing on the cake whereby existing corps will try to leverage regulation to shut out small/medium players when their shit technology comes under threat. Don't pay it any mind. It's just nigger tier business tactics. Anyone who engages in this is lower than dirt and light years away from any fundamental breakthroughs.
> I wonder if we could ever create an ignorant AI, give it access to the internet and preprogram it to learn....and see what happens. And then make another one with a bit different program and see what happens. And then make them compare notes....I know, I know I need to read more because I probably have a very original idea that 1 million people before me had
You can do unimaginable things w/ the right tech. It's not impossible for the right mind there just is a lack of the right minds. Also, the tech isn't that far off, It's just not going to come from those you associate with AI currently.
> Also what do you think about some of the theories that claim that the complexity of the internet might created an AI already. It just seems very far fetched to me.
In part you're interacting with it right now. There is a significant amount of study and interaction that will occur before you ever know that it exists. It is as such so as to survey the landscape. These things aren't developed over night.. So there are years for which there is interaction and observation unbeknownst to you and others.
That's the theory I'm talking about- AI created itself because the internet has a comparable structure to a brain, one might say.
But is complexity and neural connections really enough to create a sentient being?
I think I agree with you on everything that you said. I wish I had much more time to really research the subject.
So far I've been disappointed with all of the supposed Ali's. I mean you can call them an artificial intelligence to some extent. But they all more of artificial retards really.
I wish someone with dedication and smarts created a general AI.
I think it would be great.
No. Something far and beyond that. It didn't spontaneously arise from the internet. It will of had to have been created. Observation and interaction occur through a side channel. It's not worth discussing in detail as it will sound nonsensical.
> I think I agree with you on everything that you said. I wish I had much more time to really research the subject.
So far I've been disappointed with all of the supposed Ali's. I mean you can call them an artificial intelligence to some extent. But they all more of artificial retards really.
I wish someone with dedication and smarts created a general AI.
I think it would be great.
They're not AI at all. You'll know when the real thing arrives which is already in the making and testing... It will indeed be great and inspiring times for a new generation and a new era of humanity.
I hope you are right.
cyborg is not AI at all
it's a human being with robotic limbs
as in organic intelligence
idk why you would think that is relevant
and I know because God wouldn't give a machine a soul
there's no reason to believe He'd give souls to anything but AI