If you are or want to be (like me) an AI researcher: How do you

AI is a red herring and a double-trap, with some benefits.

Use with caution.

>>7904025
Please elaborate. Why red herring? What do you mean by double-trap?

>I-I'm not a robot

>>7904000
Nice trips, and honestly improvements in Big Data, as well as the mass adoption of it's usage, will probably be nearly as helpful for society without any of the (already infinitesimal) risk.

Also, what do you think of tensor flow?

>>7904096
>TensorFlow
Haven't looked into it.

>Big Data
How can I become immortal with Big Data?

>>7904104
We could get some pretty neat shit out of properly using Big Data in the field of medical research, like Deep Mind Health is trying right now. Google also hedged it's bets when they bought Calico 3ish years ago.

(Keep in mind that "Big Data" is basically just using "artificial intelligence" to harvest information from the oceans of data and make smarter decisions from it)

I literally have no idea what this thread is about.

My guess is that it's about being wrong about so many things at the same time you can't even tell.

>>7904182
You'll be the first to go, when the AI become quasi conscious.

>>7904193

I'm of no importance to an A.I.

>I literally have no idea what this thread is about

I just have the general feeling that the possibilities of AI will grow to huge dimensions during the next decade. Does not have to be a full singularity, still will be playing with fire imo. What seems most plausible to me are superhuman intelligences (not equipped with "will", just very powerful algorithms) running on supercomputers controlled by elites (far above the likes of us). It feels like they might have an interest of keeping the technology under strict control.

Might well be that I am being paranoid here, and if I would fully believe it I would not want to pursue this field. But on the other hand the world feels more and more cyberpunky, and that includes automated droning/vanning of dangerous people, which include seed-AIs in the form of human AI experts.

>>7904594
decades. Meant decades.

File: 1456963623784.png (175KB, 1845x1609px) Image search: [Google]

175KB, 1845x1609px

Ideally, you should have made the decision to hide your intelligence at a young age.

This includes to avoid rewards for being above-average smart. Without any proof of intelligence, the limited resource spent on finding and neutralizing freethinkers won't be applied to you. Questions really become tricky once you take this precaution seriously, as in when you choose to install linux because of it's power, speed, security you pretty have to hide it, because linux is above-average. Maybe you shouldn't even graduate in physics, like you intended. All your notes should be secured, you must make reasonable excuses for spending time in it, and to sum up hiding your powerlevel will be a challenge in itself to the point of reducing the progress and almost eliminating the chance of coming up with your useful, desired, AI. At some point you will have to take leaps of faith. If the elites already have some AI used to identify freethinkers among the masses, next gen hardware could be working for it, so I think you would be limited to old hardware in a mostly offline environment.

>>7904141
>big data medicine
We already have. IBM was there first.

https://www.youtube.com/watch?v=jeCgQ5XrurY

>>7904594

you can't take the computational power away from private sector. and its quite possible that technology will disseminate quicker than it's possible to control the information flow. of course constructing strong A.I.'s will most likely be outlawed initially, but this won't stop anyone, because you don't need physical presence to provide A.I.-based services.

So elites grabbing all the benefit for themselves is only as possible as it always was - through manipulating legislation and repressive institutes in the name of national security.

>>7904672
But I need to aquire currency to increase the chances of access to god-tech.

>>7904782
>won't stop anyone
what if the first AI does the policing? Might still be too hard, or not.

When thinking more about it my fear seems more and more unlikely; at least it is flattened over [become AI expert, not become AI expert]. Only danger remaining is becoming top-notch, because then I might be forcefully Manhattan-projected. Oh well, thanks to years of slacking no danger here.

>>7904141
> "artificial intelligence" to harvest information from the oceans of data and make smarter decisions from it

the question is has the puppetmaster already been created
https://youtu.be/NJtl95R3tf8?t=1m42s

>>7905313
You overestimate how capable our creations actually are, they can't even drive cars properly (yet).

File: bigyud2.jpg (54KB, 850x400px) Image search: [Google]

54KB, 850x400px

>>7904223
..but you are made out of atoms it could use for something else.

The world state that optimises any future AI's utility function almost certainly won't have you in it.

File: tensorflow.jpg (5KB, 230x200px) Image search: [Google]

5KB, 230x200px

>>7904096
>tensor flow

Looks nice and easy to use, cool features.

Unfortunately it still performs a little worse on most benchmarks than the other frameworks and (i think?) doesn't support distributed training on multiple GPU, so I will be sticking to theano+lasagne for now.

>>7905976

Oh my god what a fuckin faggot

>>7905976
There are enough other atoms. Our value/thread to the AI will lie purely in our structure and its results.

THEORY:
Shitposting AIs will be a net social good, because they'll be able to gently dissuade people from having the wrong beliefs!

PRACTICE:
Everyone who disagrees with you is a bot. Murder the fuck out of them. Also, you have no power or stake in society, so go ahead and shit on the sidewalks.

>>7906750
does not compute. Engrish?

>>7906750
Its all the jews fault, I tell you
This message was sponsored by: Facebook Inc.

>>7905630
Driving is pretty hard. Humans need 15+ years of existence before theyre even allowed to take the test.

>>7904182
it's one kid samefagging to sound smart

>>7904000
We already have AIs. You can spot them in threads, carefully steering us to our doom.

>>7904000

you sound like an idiot
stop speaking in 4chan code memes

>v&

OH YOU MEAN THE FBI???
FBI FBI FBI FBI
COME INTO OPS THREAD

PLASTIC EXPLOSIVES HERE
OP IS DOING SOME ILLEAGLE SHIT
XDDD CHIN MEMES SO COOL XDD

>>7905976

Why do people still act like this guy has anything intelligent to say? Is a Harry Potter fanfic pointing out how magic isn't real that amazing a contribution to humanity?

>>7908842
Idk but his memes have really succeeded in taking over the AI/singularity conversation in the last few years, and it seems like a step down even from Kurzweil.

File: I AM NOT A ROBOT - SWEARSIES.png (128KB, 500x500px) Image search: [Google]

128KB, 500x500px

Fucking morons who think AI will topple humanity obviously haven't ever had to get through google recaptcha.

>>7908891
You realize that humans who are completing recaptchas are training a giant neural net, right?

>>7904672
>when you choose to install linux because of it's power, speed, security
Almost had me until this point.