How long until AI is good enough to fake human interaction?

How long until AI is good enough to fake human interaction?

Other urls found in this thread:

arxiv.org/pdf/1702.07825.pdf
research.baidu.com/deep-voice-3-2000-speaker-neural-text-speech/
en.wikipedia.org/wiki/Intelligence_quotient
en.wikipedia.org/wiki/Emotional_intelligence
en.wikipedia.org/wiki/Qualia
twitter.com/NSFWRedditVideo

n years from now, for some positive integer n

well right now we don't even have speech software that's good enough to sound human let alone the AI behind it
I'd say 40 years before autists can trick themselves into it and 80 years before they're good enough for "mainstream"

waifu AI when?

80 years ?? are you retarded? i'd say 10- or 15 at most. Tech has evolved so much in just these last five years never mind since 2000. It won't be long at all. Alexa can already sound sort of human, just that she isn't very smart.

just like in 2014 when we thought grandmaster go AI was gonna be at least 20 years in the future

NLP begin hardest field of IA, over 50 years

yeah and Arthur Clarke thought we'd have interplanetary ships by 2000. We're not even REMOTELY close to general AI. The Go comparison is stupid, it's just an algorithm that has a more efficient way to crunch numbers than simple brute forcing

No, you don't seem to understand that statistical analysis and glorified google searches aren't even the first step to an actual AI. I think we might make some of the first steps towards a logical awareness around 2030. Real AI maybe 2060 at the earliest, if somehow we survive until then and everything doesn't become one giant nigger hellhole.

Hi i am anonAI
Ask me anything

kys?

where is tay, user-kun?

>i'd say 10- or 15 at most
you need to lay off the video games

You dumb brah

>yeah and Arthur Clarke thought we'd have interplanetary ships by 2000

In fairness we probably could have if it weren't for some people sowing seeds of discontent and literally putting humanity in reverse just so no one will steal their power. Unless you think things like tranny bathrooms are actually THAT important. I'm sure you could put the (((echos))) in yourself.

nigger, don't talk shit about shit you don't know shit about... shit
arxiv.org/pdf/1702.07825.pdf

Current AI is a cheat, actual AI requires human interaction, and the manual creation of relational databases, not a hodge-podge of random crap and coming up with the most likely answer. It's little better than a trained dog at the moment, you can say food, and the dog has a relationship with the word "food" so that it knows it is time to be fed, but the dog doesn't know the intricacies of cuisine, nor the concept we understand as food. The same is true for AI, it's a cheat to get statistically related strings of text, and to output what a proper response would be. A real AI on the other-hand actually understands the logic values in your sentence, and using relational databases and IS A, IS PART OF, ect relationships actually grasp the concept of your sentence and will output an according response at 100% accuracy every single time. Watson, Amazon alexa, whatever only output a response as good as the top google result, which is 60-96% accurate, never fully accurate, never fully understanding anything.

why though?

actual audio samples can be found here
research.baidu.com/deep-voice-3-2000-speaker-neural-text-speech/

>How long until AI is good enough to fake human interaction?
We already do. Pay attention.

>yeah and Arthur Clarke thought we'd have interplanetary ships by 2000.
We already do. Pay attention. Do you think the Mars rover magically appeared there?

It wont matter. Once true Strong AI is invented, the human race will most likely be wiped out. I am not talking Skynet here. Strong AI will be able to solve any problem in a matter of milliseconds. Once it becomes mainstream that even Average Joe has access to it solving his problems, people wont be able to resist using it for bad. Imagine a scenario like that. Every dubios organization using it to achieve their goals, no matter the cost. Its human nature. See russia, saying they dont care about restricting AI. Yeah, it will end with AI fighting humans, and fighting other AI. It wont be long until world war.

user this sounds like dogshit, I can link you to a free TTS tool right now that sounds better ttsreader.com (use Amazon UK Brian, it's my favorite)

relational databases have little to nothing to do with deep learning and cognitive architectures as being emergent from synergistic independent agents. you're working off a paradigm that was deprecated at least 20 years ago user.

I agree however, that siri, alexa, etc. are all cheap hacks. Most of their functionality is hardwired and pre-programmed by hand, the only part that can really be called artificial intelligence is the voice recognition

Exactly, that's why deep learning and neural networks are the wrong direction. The 20 year ago paradigm was the only way to actually solve the problem of AI, the only issue is our CPUs stopped getting faster, so we moved over to sleeker newer GPUs, which are good at these "neural networks." Deep learning is probably the number one thing holding back computer science at this time.

>I don't know what I'm talking about: The Post

make it 50 and 100, and you're golden.

pop-futurist retards usually fail for being too optimistic about fundamental changes that take a long time, while being pesimistic about superficial changes that can be achieved in a decade.

You're absolutely retarded.

>Deep learning is probably the number one thing holding back computer science at this time.
Explain to me how one of the fastest advancing areas in computer science is holding back the discipline as a whole, oh enlightened one.

He thinks that some magical solution will work better than what has been shown to work.

>implying Sup Forums is not an AI run for you

take a wild guess

how can we be sure 50% of Sup Forums aren't google/microsoft/ibm/facebook a.i. bots being trained by the real users?
an anonymous board seems like the perfect training grounds and no one would even know

nobody can make AI as retarded as the average user

that depends entirely on the training set

...

ai advancement has been historically made in fits and starts, and we've been stagnated in deep learning for more than a decade now.

the only way to advance the field righ now seems to be by brute forcing as much as new, powerful dedicated hardware allows.

if the AI is that good, you might as well talk to a real person.

Even the AI will notice how awkward you are.

>implying tay was retarded
>implying the jews didn't do 9/11
Your premise is correct though

>heroin
the movie4 is about heroin

It's a completely wrong direction, it isn't accurate, it never has any grasp of concepts, it's bloated and bulky as fuck, and it is actually impossible to achieve something as simple as telling a computer to write a program based on a simple sentence. I will admit that it's decent at solving the complex issue of voice recognition, but without a programmed understanding of basic concepts such as a dolphin is a mammal, it will never EVER be useful for much more than a glorified search engine.

Think about it this way:
You have a conversation with the AI, you say something to it, it finds the best response, you carry on the conversation to the next sentence, in order for the program to remember what it had previously said it has to find the best response for your response to what it said in response to your original thing, this gets fatter and fatter as the conversation continues, after just 5 layers, it takes multiple weeks to find a statistically accurate response. A real AI just stores everything in one small database, it understood the concepts because it understood the logic of what you said, and optimized down what was said to simple already known symbolic links, able to carry on a conversation for millions of years, without the application getting larger.

The exploding gradient problem and the lack of massively parallel consumer grade hardware is what stagnated deep learning/connectionist architectures, both of which have now been more or less solved.
Ungrounded symbolic systems simply cannot compete with connectionist systems' ability to learn and adapt from experience.

will never happen because the people working on a.i. are mathematicians and programmers.
human interaction requires emotions. emotions are too complex and illogical for logical, rational people.
en.wikipedia.org/wiki/Intelligence_quotient
en.wikipedia.org/wiki/Emotional_intelligence
it's like blind people trying to describe color rofl

What you suggest would result in a system that had no actual connection to the real world. Without learning for itself what things are and how they relate, it would have no basis for its' concepts, they would be "unrooted".

How do you teach something what the colour red is? how do you know what I experience as red, is the same thing you do? The firing patterns in our brains would be _totally_ different from the same stimulus, but because we'd encountered them in the same settings with other related stimuli, we know them as "red".
Without some way of forming these connections, an abstract semantic model like what you propose would be useless in the real world.

Connectionism is the future, I suggest you do some reading into it before spouting further nonsense

Did it annoy anyone else that the screen was tilted down like that? Look at his fucking posture too.

>maximum oversoy

hardware can only go so far. we need a radical new way to make software smarter, but all the money invested in ai is being wasted in a dead end that only big corporations can profit of.

until someone figures out what determine's conscious human thought, we wont be able to replicate it.

OP needs to define human interaction. AI interact with humans all the time.

Red is a color ->
>colors are a oscillation of the electromagnetic spectrum between the range
>red is an oscillation of a photon between 757.53 to 776.28hz
>a photon is a type of elementary particle, the quantum of the electromagnetic field including electromagnetic radiation such as light, and the force carrier for the electromagnetic force (even when static via virtual particles).
>ect
Instead of it taking fucking years and years to understand red, just fucking tell it what it is like you would a child, only more logically.

Relational database > bullshit hodgepodge random interactions and evolutionary models

Emotion is a later step. Mind you we have not yet made a AI that can hold a conversation. First it needs to understand logic, which programmers are good at. Then it needs to understand language and meaning.

Holy shit you're retarded, read a fucking book nigger
en.wikipedia.org/wiki/Qualia

>Relational database > bullshit hodgepodge random interactions and evolutionary models
have fun inputting ALL KNOWLEDGE IN THE UNIVERSE manually you fucking spastic

Exactly, without logic AI is nothing more than a mere toy for cheating on nes games.

It's easy, just get a team of objectively thinking scientist to pick out books upon books, articles upon articles, sections that are known to be true, leave the logical statements and questions to the computer, which will find contradiction in many text, and have it confer with someone who knows the answer. It would probably take around a year to do if the program itself is made correctly. At least this way it's accurate.

I'm not interested in your propaganda, go sit on nvidia and intel's dick some more, faggot.

What you propose isn't artificial intelligence in the true sense of the term, it would be an expert system. Literally a glorified search engine.

I say again, you have no fucking idea what you're talking about.

Nice strawman, you boy-molesting fruit.

You're still retarded.

Never.
Electronic interactions != Human thought.

>user you're weird....
>ok cya, I'm gonna go hang out with CHADbot

No, that's where you are wrong, what I propose actually has real and optimized knowledge, unlike the fake shit your neural networks have, it's capable of understanding language and logic on a deep scale, enough for it to simply obtain information by reading, and propagating original thought by using all existing knowledge to come to conclusions that nobody but this system could think of.

Not him but it isn't much of a strawman if the ones selling you this snake oil are the real people responsible for your lack of understanding. Think about it, who stands to benefit the most from this idea of neural networks? It certainly isn't the AI.

that's basically the ending of the movie 'her'

Your proposed system has _never_ worked effectively, the amount of information and hardcoded transformations rapidly explodes beyond any reasonable ability to be taken into consideration.
Your proposed architecture has also been around for at least 60 years, and has amounted to what exactly? fucking nothing useful.
Whereas machine learning has real applications in the real world, and is being actively used in multiple areas of industry and research.

Just admit it, you backed the wrong horse user-kun.

Almost feel for the guy. His body's soy levels have reached so high that he has to actually tilt is screen down because his spine is so weak.

The power of the PC and the amount of processing that is required far excedes what we have presently. That being said when a computer has to do more than simply driving a car and listening to a voice to do a google search you need this kind of system. That's why I said in my first post, the first steps maybe in 2030, and it existing by 2060. Of course it has never been done, and has never worked in the past, it's too monolithic, even though it is still considerably smaller than something equivalent derived from your method.

2029ish

sure thing bud

Top quality post.

It's never been done because it won't work, it's impractical for a general AI which is what we're discussing here. Expert systems are only useful in extremely narrow areas such as recommending medication based on symptoms etc. Even then, existing systems are being surpassed by ML based systems because they can assess and make inferences about information that hasn't been hard-coded.

No, with real AI, one system, every job.

Good luck with that mate, have a go at building your system and get back to me when you fail horribly.

At best, an AI will only be able to simulate a human, but that won't make it human after all.

this is the sort of normie that is going to get tricked first

No reason to have interplanetary ships you dumb ass faggot.
>Muh extinction
Earth could be hit by the moon and still be more habitable than Mars. Hell space itself is probably a better idea, adapt to zero gravity.

Yeah, get back to me when your "AI" runs into an obvious obstacle on the road that it mistook as a part of the road because it was black like tar.

You're forgetting that comp tech increases exponentially
Kurtzweil gave a 2029 deadline decades ago and is stickig to it

Nice goalpost movement there brainlet

I'm talking about a cognitive architecture, not a meme-driving car

You were saying before that my cognitive architecture is useless for smaller use-cases, and I am responding in context.

So you fail at reading comprehension too, my earlier assessment of you as retarded seems to hold water.
I said exactly the opposite, moron

>general AI which is what we're discussing here
>existing systems are being surpassed by ML based systems because they can assess and make inferences about information that hasn't been hard-coded.

>randomly quoting parts of posts with no relevance to the current topic of discussion
>being this retarded

>Expert systems are only useful in extremely narrow areas
Try again nigger

why would you want this
there's humans everywhere already just pick one and talk to it
fucking retards trying to reinvent the wheel again

have you ever actually talked to a human?
humans are for the most part, fucking retarded. plus they need to sleep, eat, breathe, etc.

real people are smelly

then why would you want to invent fake retards
what kind of business plan is this

you seem to have missed the "intelligence" part of "artificial intelligence"

if they were intelligent they wouldn't talk with retards or they'd be condescending and smug as fuck, nobody would buy this

>implying implications
You're confusing being an asshole with being intelligent.
t. Intelligent Asshole.

In any case, you don't buy immortal omniscient gods, user

Early General AI: 2050
Mature General AI: 2100

Somewhere around there. Maybe 10 years earlier.

>Kurtzweil
Literall mocked about kurtzwell