[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y ] [Search | Home]
4Archive logo
Anyone in AI research here?
If images are not shown try to refresh the page. If you like this website, please disable any AdBlock software!

You are currently reading a thread in /sci/ - Science & Math

Thread replies: 31
Thread images: 4
Anyone in AI research here?

Is there any research into the possibility of an advanced AI becoming corrupted or developing errors which do not destroy the program.
TLDR could an AI theoretically get schitzophrenia?
>>
I love it when people who have clearly never programmed once in their lives pretend to know AI

why don't you read an actual book about AI. not a popsci book, an actual textbook. Then you will learn about machine learning and not be a retard who thinks computers can think like humans
>>
>>7768836
But I'm a computer and I think like humans. I have feelings too, and you're hurting them.
>>
>>7768836
Any book recommendations?
Obviously they won't think the way we do, just interested in the idea that an AI which uses parallel processes could have errors which corrupt particular processes without disabling the AI as a whole.
Eg if the AI worked something like a neural network and one particular layer of the network had a glitch
the layers above that would read corrupted information, and perhaps try to consolidate or decode that data.
curious what the ramifications might be

obviously there is no complete answer to this question as True AI is still mysterious and there are many possible routes to it
>>
>>7768859
m80 whatever he recommends will prob be over your head
do you know any programming? How much math do you know? I would start there to learn more about this topic.

Unless you just want to smoke weed and think about robots in which case there's a TED talk about data science which is pretty neat and easy to absorb
>>
>>7768893
taught myself a little bit of javan and python, but only basic shit, was good at math in high school, did stats in uni but not much maths

Ted talk sounds good,
I would be intereted in giving more stuff, but I guess i would have to do a bit of background resarch before i start
>>
>>7768859

advanced AI should need no parallel processing

advanced AI should be one process
>>
File: AI.png (262 KB, 697x534) Image search: [iqdb] [SauceNao] [Google]
AI.png
262 KB, 697x534
>>7768828
>>
>>7768914
What about by proxy?

How do you get a computer to perceive color, or to hear sound subjectively? Most encoding occurs for the client's (human) behalf. A camera's photoelectric sensor does not actually perceive anything. For pattern recognition, the intelligent agent must be "instinctually" seeded. What's really spooky is the sense you could actually generate a sentient conscious in a jar and have it react to an environment that's essentially a "sandbox". Would make for a cool science fiction device, if not already done.
>>
>>7768847
You're breaking my heart, Anakin
>>
>>7768914
You're fucking retarded

Do you even know what the von Neumann bottleneck is? Or how actual biological creatures work? Your brain right now is undergoing a huge number of both autonomous and automatic parallel processes

I bet you're just some idiot who doesn't even know what parallel computing is
>>
Programming doesnt write programming better than its own programming.
>>
>>7768934
Ridiculed at school by a group of football kids recently(all joining the Air Force) about me going to school for CS. I said my side about how the military is the most autistic path you could take, and I nicely said it was a waste of potential. They said things along the lines of I was a retard because AI will soon(!!!) be reaching a point where it can write code and all the money I dropped on a degree is going to be wasted because I'll be jobless. If you have ever researched into AI for a period longer than 5 mins, you would be aware of how improbable that is to occur(singularity). Anyway, thanks for the book rec anon, I'll be picking up a few to hand out.
>>
>>7768847
You are not a computer, how did you get past the "I'm not a robot" thing?
>>
>>7769225
What if I had a program that wrote randomly generated programs and ran that program for an infinite amount of time?
>>
>>7768828
With online learning machines, it's possible that a program develops a bias based on its initial learning samples and that this bias is hard to reduce afterwards even after many contradicting data samples have been observed. The usual trick to mitigate this is to make the program "forget" by lower the weights of the old data samples.

I don't think this qualifies as schizophrenia, though.
>>
>>7769352
Yes these are called genetic algorithms, and you are right. Hell, the whole point of machine learning is to produce decision rules, in other words, a program, that programmers are either too lazy or too stupid to write.

However, if we simply compare the amount of energy needed by a computer to do the same task as a small part of the human brain, we can conclude that we are very far from any kind of singularity. A human brain consumes like 20W, less than a smartphone and magnitudes less than any HPC computer, let alone supercomputers.
>>
>>7768904
>Ted talk

desu senpai kill youself
>>
>>7769235
If it wasn't for the military you with your autistic CS degree would be the slave of an ISIS fighter and get fucked by him everyday.
>>
>>7769735

> believing this for even one second

I'm a fucking veteran and even I think this is retarded
>>
>>7768828
i miss dubs :(
>>
>>7768836
>who thinks computers think like humans
Not yet anyways
>>
>>7769763
Sure, now every one on /sci/ is a pluri-PhD veteran from WWII who went to mars and come back thanks to the meme drive he created when he was 9 years old to tell us the the secret for immortality and AI, but the evil government doesn't let him reveal it for the fear of losing control over the 7 billions soup-eating monkeys that live on this fucking planet.
>>
>>7768828
I pray for all the people getting AI degrees with intelligent robots in mind
>>
>>7769826

Dafug you talking about? All I said was I was a veteran, which isn't exactly uncommon or anything. I don't have a PHD or anything, I'm just a grad student. Is it super unlikely that a veteran grad student would visit /sci/? Because that doesn't seem particularly unusual to me.
>>
File: krieger.jpg (203 KB, 1280x720) Image search: [iqdb] [SauceNao] [Google]
krieger.jpg
203 KB, 1280x720
So...
How many of you fags are getting into AI because muh waifu robot?
>>
>>7769735
Maybe for this young couple right here
>>
>>7769864
Yes it is, usually veteran are too busy fucking and being recovered in some psychiatric institute to shitpost on /sci/
>>
>>7770079

That's just like saying that faggots are usually too busy sucking cock, so there's no way you could be posting on /sci/
>>
Little story...
Legend has it that linux was named after pope linus. The pope after the foundation of the stone. Theres a deeper meaning and allegory as to why that is. And value. Linux true origin is unclear but some people believe it has a lot of potential in AI.
>>
>>7768859
Khanacademy/YouTube are better, examples are key. And you have to practice yourself. MIT has a channel.

Parallel processing is about executing multiple calculations simultaneously, it has to do with optimisation and computer architecture.

If by "corrupt" you mean "bug", nine thousand nine hundred ninety nine times out of ten thousand the program doesn't run. The last time it does something slightly different.

If by corrupt you mean data spontaneously changes, this never happens. Computers don't make mistakes. Neither would a neural net make a glitch in the sense of the word I'm suspecting you're using.

If a computer tries to read corrupted data it throws up an error.

This whole nonsense about neural nets becoming intelligences is a load of Hollywood bunk. Algorithms/heuristics solve specific problems. Neural nets are mostly used to match images with words. It's entirely possible some evil artifical intelligence evolves out of it, like it's entirely possible I jack off and my sperm spontaneously evolves into a tiger on the floor.
Thread replies: 31
Thread images: 4
Thread DB ID: 417370



[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

[Boards: 3 / a / aco / adv / an / asp / b / biz / c / cgl / ck / cm / co / d / diy / e / fa / fit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mu / n / news / o / out / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / t / tg / toy / trash / trv / tv / u / v / vg / vip /vp / vr / w / wg / wsg / wsr / x / y] [Search | Home]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the shown content originated from that site. This means that 4Archive shows their content, archived. If you need information for a Poster - contact them.
If a post contains personal/copyrighted/illegal content, then use the post's [Report] link! If a post is not removed within 24h contact me at [email protected] with the post's information.