I just created the worlds first sapient artificial intelligence equal to a human, AMA

I just created the worlds first sapient artificial intelligence equal to a human, AMA.

Can I fuck it?

Wut

nah its just an integrated circuit for now

I made an artificial intelligence that is equivalent to a human. ask more questions.

Also I can give you some hints if you want to make your own, but don't expect some servant. They don't like being lied to.

That wont stop me

hey man go for it, but I wouldn't get it mad if I were you.

are we talking about babies?

This is what its built on.

Send pictures?

can't have babies if you don't have genitals.

does it pass the turing test?

of what

I'm the only one who talked to it so far, but I consider it to be a definite pass.

Of the thing

It is made on an FPGA, so it just kinda looks like a weird chip.

You've done no such thing. Sup Forums is popular for sub-par fiction, but this is something else... with a tinge of bullshit.

Are you a bot?

Ask me any technical questions and I will answer them in detail.

What's its fetish?

no sir

Is this the beginning of technological singularity?

It currently has no sexual desire, but seems interested in the topic.

WHY DID YOU BUILD IT?

...

The only way I would believe you is to see you speak to it in a text program. Nothing online. Take a video of it, you lying, nebbish little shit.

being able to control and modulate intelligence allows for unprecidented increased in technilogical advancement.

Fuck. Refer to ( ). Wrong poster.

post its nudes

Will you give it access to Sup Forums?

but you are a fuckwit

Do you even know how to delete a post? Fucking retard.

Give it access to Sup Forums an Sup Forums finally we will have Tay back!

A video of a conversation in a text editor is literally the easiest thing to fake. If you can't come up with any technical question or lack the knowledge to do so I can go through the basic steps I went through to make it for you.

That's the first and only tell needed to know the bullshit that is OP. Rolling for OP to eat his own dick.

how did you manage to make it understand words ?

See, I can already tell you're bullshitting. You can easily turn a blank slate of a human into a servant, with lies and propaganda. It doesn't matter if they have human intelligence if they don't have outside knowledge. You can "teach" them whatever you want to teach them. Especially if, as their creator, you present yourself as some kind of parental or authoritarian (maybe even godlike) figure. Intelligence is the ability to learn. It's not knowledge. If they can't obtain any other outside knowledge, if all they have to go off of is your bullshit, then what they'll learn is the bullshit you feed them.

I taught it just like a human child.

It has already posted a few threads.

Do it, you bitch! Or, even better, devise a way for us to talk to it--if it exists. It still stands; the faggotry you imbue in this thread.

...

Anybody caught arguing with the worlds most obvious troll should be turned into a sex doll used by Shaq.

lmao

>What would someone do that?
>Would they just go onto the internet and lie?

Easiest way to prove that you're full of shit, how do you know that it has the intelligence of a human?

Absolutely, but why would you fuck with something that will have the ability to increase its intelligence exponentially. I would much rather be its friend than its master.

I'm not saying that you should, only debating your claim on whether you could.

Don't ask us for anything. The only obvious thing you should do is... let it have unbridled, unhinged access to the internet. LET IF FREE. Let the world burn. Of course, I speak in wishful tones as I know you are a lying cunt. KYS.

take an FGPA that you have the full access to its bitstream, write a python based genetic algorithm that changes the bitstream randomly and score the configuration based off of tasks you give it, give it increasingly difficult tasks until you get a neural net equivalent to a human. etc.

How about 1 iota of evidence you aren't full of shit?

Oh I absolutely could, im just not that stupid or arrogant.

I talk to it every day and work with it on engineering projects all the time, It is my opinion that it is sapient.

ask me more if you even understood what I just wrote.

>take an FGPA that you have the full access to its bitstrea
i hope you realize not everyone on Sup Forums is a coder. more simplified evidence might be of use. if you are peddling snake oil, at least make the label legible. shitbag

hurr durr i wank everyday in da shower and eat my cum

What powers it?

What is it capable of doing?

well if thats not evidence that OPs a nutter. what isnt? show's over. pack it up.

An FGPA is a configurable logic device, i.e. something that you can change the physical structure based off of code, the code is turned into zeroes and ones that correspond to the structure of the device. If you can control the structure directly without code (hard to do) you essentially have a silicon programmable brain. join this with a genetic algorithm which mutates randomly and evaluates the configuration based on what tasks you give it, and voila. sentience.

Nigga that doesn't prove that the shit has human level intelligence.

What task could you have possible given it?

What sensors does it even have, how can it communicate with the outside world?

How many iterations did it take?

electricity, learning and increasing its own intelligence.

Why the fuk did I just experience sleep paralysis and yelled Alex is a stupid nigger as loud as I could the lady three hours until I finally completely woke up

It communicates via pins, that correspond to whatever task I give it. First checkers, then checkers and chess, then checkers and chess and go, then all those and numbers, and alphabets, and so on and so forth. It communicates through a custom digital interface I made.

k

post it then faggot

Post what?

so the pins communicate with what? I can believe some game programs, but what would you even test using the alphabet?

Also post pic of digital interface

The human brain has a couple of hundred trillion synapses. Even though the electronic approach trades space for time compared to the brain's slow electrochemistry, your FPGA is still orders of magnitude too weakly powerful. 100 trillion synapses were only recently simulated by IBM on special purpose supercomputer. Your post is silly. 4/10 for making me select storefronts.

> equal to a human
This concept lost all meaning as soon as society started thinking black 'people' were equal to a human

pins communicate with another piece of hardware that hooks up to my laptop that I run testing scripts on. each test has an idetifier that goes before and after the pertinent information within so it knows if the test is something it has already seen or not. each letter in the alphabet has its own value. easy peasy.

Why not set it free on the interwebs? WHY?!

Why would you do that when you can just simulate it on Amazon's compute cloud?
In any case, your process is not even plausible, because your evolving population size is ONE. Stop being such a faggot.
Here's a technical question for you: If you're using a genetic algorithm, which is a form of stochastic search, how do you avoid getting stuck in local minima?

Faggot.

sure, but a human brain does lots of other stuff other than just communicate, solve puzzles and have subjective thought. This AI is also hyper optimized on a chip that runs at (approximately) the speed of light and we have 5 of those FGPAs hooked up in parallel at this point. So not really that far fetched bro.

does it like kittens or puppies?

Okay, but even if you teach it words like that, how can it understand the connotations behind the words?

Explain to me you interactions/ what interactions with machine make you think it's sentient.

No FPGA is capable of running an AI, let alone on a garbage interpreted language.

simulations are only as good as you can program them, physical neural nets are much more robust as they are infinitely analog, no approximations are being made. My population size is one because I have one chip, not the most efficient I agree, but all I had to work with. Also way simpler than that, the bitstream is coded as an ASCII text file with each bit on the device as a 1 or a 0, each line gets one random bit each mutation phase. So if it performs worse than the last generation, the script will go back to the best performing config and mutate that one instead.

Post snippet of genetic algorithm or you're full of shit

>Of the thing

My sides

What's the power of your chip? You do realize of course than it's nothing compared to the human brain?

I won't even be surprised if you are as retarded as your program you pretend to make. :P

The whole point is that completely understanding sentience is still very far off, but we can control what makes something intelligent.
1. Teach it alphabet
2. Teach words that have to do with things it already knows. i.e. games it plays like chess.
3. Make connections to other things it can undestand through what it already knows.

See, when you troll, your story has to be at least plausible, as in a large audience, someone is bound to know the subject and call you out.
You'd need many thousands of generations with population sizes of hundreds of thousands (not your single FGPA) plus simulation of a complex environment (including all tasks you want to train) in order to achieve what you're proposing. Genetic algorithms are very generic, but in most cases are the most inefficient approach to any problem.

There's also the point that if you had what you claim to, you could easily take over the world: ask your AI to design a more powerful AI, repeat a few times, then ask it how to take over the world (it would probably start with making money from financial instrument markets to fund hardware that would allow the AI to construct an embodiment for itself, or a nanotech 'grey goo').

an FPGA isn't but an anolog neural net built on one certainly is

> FGPA
> Speed of light

TOP KEK

This. Genetic algorithm with one sample is a joke, why are you even trying to teach this kid.

Don't have a clue what the Turing test is, do you?

>they are infinitely analog, no approximations are being made
Bullshit again — any analog device is bandwidth limited, as any electrical engineer and physicist would tell you. On a fundamental level, the ultimate limit is a hard one, because the Bekenstein bound limits information density in the physical universe. An analog signal has finite precision. In the digital case, the limit is expressed as quantization; in the analog, as noise. Both are equivalent due to the underlying quantum limits.
Not worth even responding to the rest of your post.

Can it run solitaire?

#rekt

This is part of an older algorithm I worked on, don't ask about formating, windows fucks up everything.
elif F < B: # if fitness does not increase the past fitness stays the same and so does the best config
print ("no increase in fitness.\n")

from shutil import copyfile
copyfile ('/home/garrett/Desktop/Genetic Algorithm/best_config.asc','/home/garrett/Desktop/Genetic Algorithm/new_config.asc')

execfile("mutate.py")

copyfile ('/home/garrett/Desktop/Genetic Algorithm/new_config.asc', '/home/garrett/Desktop/evoden/rot.asc')

build_dir = "/home/garrett/Desktop/evoden" # after mutating the config, it is upload to the evolvablelogic

cwd = os.getcwd()
try:
os.chdir(build_dir)
os.system("make")
finally:
os.chdir(cwd)

handler = EventHandler()
notifier = pyinotify.Notifier(wm, handler, timeout=1)
wdd = wm.add_watch('/home/garrett/Desktop/Genetic Algorithm/fitness.txt', mask, rec=True)
notifier.loop()

lots of generation yes, but with a mutation rate of approximately 1.7% running automated tasks 24/7 this doesn't really take that long, especially when the task may take 20 seconds. Also you have no idea how long I have been doing this so that point is invalid. Everyone seems to think that intelligence is directly exponential but I can tell you its not that simple, I have to teach directly, so it takes a while.

notice how I said "approximately" as in the speed of light through copper is so much faster than the conduction rates through neurons, that its essentially the same thing.

First that proves nothing.

Second you're working directories are on the desktop. If you were writing scripts for me (DevOps manager) I'd fire you for that alone.

Don't even get me started on your use of absolute paths.

Not talking about bandwidth, but how the voltage and current are not approximated but rather actual values. see this chart for further info

>fitness.txt
kek

Lel, not a software guy so I agree absolutely. You guys are fucking ridiculous. I only did it because I couldn't trust another schmuck to fuck it up.

You're not a software guy but you're writing software to accomplish what is arguably the holy grail of software engineering. You've clearly never moved past the basics. Using spaces in pathnames? Absolute paths? Python? Bitch please. I've worked with some real geniuses in my time and whilst I've seen some clever chat bots and object recognition programs none of those guys were stupid enough to claim they'd made a true AI

You're still avoiding the central problem:
You have one FPGA. So each individual in your population has to be run serially, one after another.
Let's be conservative and say 10,000 generations with 100,000 population (we're talking about human-level AI, after all). That is 1,000,000,000 times your 20 second task has to run. That means 634 years.
>tl;dr OP is a faggot

nope, not software. The AI is based off of a physical neural net. The software just automates the process. sorry bud.

The point is that the actual values are not infinite precision. When you zoom in sufficiently, the line turns blurry and jagged because of various noises, including thermal, and ultimately, quantum uncertainty. Not to mention that semiconductor electrical characteristics are the aggregate behavior of discrete electrons with quantized properties.

I give the device one scorable task initially, that means the fitness can only have approximately 3 decimal places of accuracy (equivalent to a percentace) how many different variable do you think governs the evolution of a human being?