Is it possible for a short program that has some self-teaching

*to achieve

currently most AI research require some kind of evaluative/instructive feedback. define self-teaching and we can continue discussing.

>>8480262
Well it is possible in my opinion. Just look at life itself. It started from nothing with no inteligent design whatsoever and by pure chance we got where we are now and if it's possible for us why wouldn't it be possible for some computer program.

You would have to place it into an environment that promotes this development. This matters a lot.

The simple answer: technically yes, but I deem it highely unlikely hence it will not happen.

File: AI.png (262KB, 697x534px) Image search: [Google]

262KB, 697x534px

>>8480262

>>8480283
Ok fine.

I'll read that book but is it possible for me (i have very rudimentary coding skills) to write some AI software that develops into stronger AI software (ie it get's better at learning)?

>>8480280
>It started from nothing with no inteligent design whatsoever and by pure chance we got where we are now
*tips fedora*

>>8480262
>a short program
Yes, as possible as it is for a cockroach to understand fluid mechanics and multivariable calculus.

There are size limitations.

Some companies probably have the means to achieve strong AI right now. They won't do it because they can't make it empathy isn't logical and they know they could unleash hell if not careful.

>>8480262
if you run it for ten million years and it's the right kind of thing and you're lucky

>>8480318
HAHAHAHA

>>8480393
That's not true at all. A.I. is not nearly as smart as you think it is, and won't be for a long time.

File: images.jpg (15KB, 443x332px) Image search: [Google]

15KB, 443x332px

>>8480332
>Tfw you struggled with multivariate calculus
>Tfw you can't understand fluid mechanics
>Tfw you're dumber than a

>>8480474
COckroach

Not OP, but wouldn't it be possible to use a non-linear solution method to create a system. Where-in you could give a program the ability to create its own interpretations of more smaller non-linear systems?

For example, a neutral net that is give instructional feedback at first as to how to create its own neutral nets based upon a given data set of "x". Where "x" is the most generalized version of the input you wish the system to self-teach around?

Any good /sci/ alternatives that don't allow teenagers?

>>8480810
Academia

>>8480262
Is it possible for /sci/ to not have AI threads?

>>8480789
>neutral
>lel

>>8480318
If it were possible for you, why wouldn't one of the many people smarter than you do it?

>>8480823

Sure...

What would you like to talk about Anon?

>>8480262
Nice try t800

>>8480262
Yeah if your understanding of those words can be changed.

>>8480262
>short program
as opposed to the long programs that have self-teaching abilities that have achieved strong AI?

When we program AI with its main goal being to survive and flourish, and it succeds, we'll know we've created true AI

*exhales weed*

>>8480262
VAGUE
AS
FUCK

Memory Limitations:??? not listed
Computational Limitations:??? not listed
"short": vague and subjective
has 'self-teaching' abilities: vague as fuck
don't know if you mean it has a training phase or uses unsupervised learning

>curious pop-sci fanboy
nice disclaimer