/sci/ - Science & Math » Thread #11773541

816KiB, 3265x1599, singularity.png

View Same Google iqdb SauceNAO

Anonymous Mon 08 Jun 08:29:54 2020 No.11773541 View Reply Original Report

Quoted By: >>11773556 >>11773611 >>11773623 >>11773627 >>11774414 >>11774423 >>11774452 >>11774470 >>11774678 >>11774767 >>11777999 >>11778487 >>11778824 >>11779339 >>11779601 >>11779794

What should i study to invent general artificial intelligence?

Anonymous

Anonymous Mon 08 Jun 2020 08:35:03 No.11773550 Report

Quoted By:

speleology and hotel industry

Anonymous

Anonymous Mon 08 Jun 2020 08:38:49 No.11773556 Report

Quoted By: >>11773588

>>11773541
>general artificial intelligence
>>>/x/

Anonymous

Anonymous Mon 08 Jun 2020 08:57:49 No.11773588 Report

Quoted By: >>11779554 >>11780607

>>11773556
>believing in metaphysics instead of a material basis to consciousness
>>>/x/

Anonymous

Anonymous Mon 08 Jun 2020 09:07:44 No.11773611 Report

Quoted By:

>>11773541
Gender studies and fashion design

Anonymous

Anonymous Mon 08 Jun 2020 09:18:26 No.11773623 Report

Quoted By:

>>11773541
Whatever you will think will get you the most money that you are smart enough to do. Then invest and give the money you make into AI related industries.

If you're dead set on doing it yourself than probably a data science courses. Machine learning and statistics/maths have a fairly big overlap.

Anonymous

Anonymous Mon 08 Jun 2020 09:22:41 No.11773627 Report

Quoted By: >>11774490

>>11773541
Astrology and dress making

Anonymous

Anonymous Mon 08 Jun 2020 15:46:40 No.11774414 Report

Quoted By: >>11780613

>>11773541
CS, Math, and Physics. If you truly do it all by yourself, you might have to understand some sort of wibbly wobbly quantum computing nonsense or whatever.

Anonymous

Anonymous Mon 08 Jun 2020 15:49:03 No.11774423 Report

Quoted By:

>>11773541
Riflery and survivalism for when the neoluddites try to murder you

Anonymous

Anonymous Mon 08 Jun 2020 15:57:52 No.11774452 Report

Quoted By: >>11774526

>>11773541
>innate human intellect is somehow increasing over time

what a retarded dumbfuck graph

Anonymous

Anonymous Mon 08 Jun 2020 16:03:11 No.11774470 Report

Quoted By: >>11774550

>>11773541
the singularity describes the point where we become slowly slave-cyborgs. it already begun, (micro chip implantation)

Anonymous

Anonymous Mon 08 Jun 2020 16:14:06 No.11774490 Report

Quoted By:

>>11773627
This, but ironically

Anonymous

Anonymous Mon 08 Jun 2020 16:24:24 No.11774526 Report

Quoted By: >>11775057 >>11775613

>>11774452
> Never heard of the Flynn effect

what a retarded dumbfuck person

Anonymous

Anonymous Mon 08 Jun 2020 16:33:43 No.11774550 Report

Quoted By:

>>11774470
We’re all already a slave to something. At least with a brain chip I can fuck my waifu

Anonymous

Anonymous Mon 08 Jun 2020 17:19:45 No.11774678 Report

Quoted By:

>>11773541
Brick laying and Plumbing.

Anonymous

Anonymous Mon 08 Jun 2020 17:52:28 No.11774767 Report

Quoted By: >>11775184 >>11775205 >>11775601 >>11778756 >>11779339

>>11773541
Computer Engineering. We already have the algorithms, we just need more computation.

t. AI PhD

Anonymous

Anonymous Mon 08 Jun 2020 19:25:56 No.11775057 Report

Quoted By:

>>11774526
>Flynn effect
Kek

Anonymous

Anonymous Mon 08 Jun 2020 20:00:26 No.11775184 Report

Quoted By: >>11775244 >>11775255 >>11775423

>>11774767
Didn't GPT-3 show that is not in fact the case?

Anonymous

Anonymous Mon 08 Jun 2020 20:06:06 No.11775205 Report

Quoted By:

>>11774767
>Computer Engineering
I second this. The future of AI is in hardware.

https://www.intel.com/content/www/us/en/research/neuromorphic-computing.html
https://www.youtube.com/watch?v=TetLY4gPDpo

Anonymous

Anonymous Mon 08 Jun 2020 20:18:42 No.11775244 Report

Quoted By:

>>11775184

Nah, modern models show improvements as you add more layers/parameters, with no end in sight. Microsoft is preparing a 10 trillion model.

Anonymous

Anonymous Mon 08 Jun 2020 20:21:41 No.11775255 Report

Quoted By:

>>11775184
GPT-3 is not sentient or even protosentient, but it is suggesting a path to broadly useful ai structures.

Anonymous

Anonymous Mon 08 Jun 2020 21:07:33 No.11775423 Report

Quoted By: >>11775521

>>11775184
GPT-3 is limited by the (self-)supervised learning approach. Reinforcement learning agents with models this large (and larger) will eventually lead to better results. However, these are less sample efficient and so need more time to reach the same level; all the more reason to make computers faster. By the way, it’s probable that we’ll invent better algorithms, but this too is driven mainly by computation: with more computation we can run more experiments and therefore home in on the best algorithms faster. The whole reason AI is hyped right now is because around 2009 we realized that with modern computers artificial neural networks are actually useful, but artificial neural networks were already around in the 40s.

Anonymous

Anonymous Mon 08 Jun 2020 21:39:58 No.11775521 Report

Quoted By:

>>11775423
Another thing: AI used to be quite theoretical, involving lots of math and proofs. Nowadays it’s almost entirely empirical/experimental. Any proofs almost always depend on assumptions that don’t hold in the real world. Nobody really knows why methods work. We just throw shit with lots of computing power at the wall and see what sticks.

Anonymous

Anonymous Mon 08 Jun 2020 22:09:08 No.11775601 Report

Quoted By: >>11775683

>>11774767
how important is mathematics and statistics to the AI field?

Anonymous

Anonymous Mon 08 Jun 2020 22:14:17 No.11775613 Report

Quoted By:

>>11774526
The Flynn Effect has been going in reverse for years now

Anonymous

Anonymous Mon 08 Jun 2020 22:36:55 No.11775683 Report

Quoted By:

>>11775601
Very important, but nowadays you can get away with a rather shallow understanding. Software (e.g., PyTorch) does most of the heavy lifting.

Anonymous

Anonymous Tue 09 Jun 2020 09:12:37 No.11777472 Report

Quoted By: >>11777998

Do AI and GAI need knowledge of the same filed?

Anonymous

Anonymous Tue 09 Jun 2020 13:28:36 No.11777998 Report

Quoted By:

>>11777472
GAI is the ultimate ambition of AI, so yes.

Anonymous

Anonymous Tue 09 Jun 2020 13:29:25 No.11777999 Report

Quoted By:

>>11773541
do you have something better to do?

Anonymous

Anonymous Tue 09 Jun 2020 16:14:43 No.11778487 Report

Quoted By:

>>11773541
that's not what a singularity means dense retard, if you can't understand a simple mathematical term you sure won't invent fuckin agi

Anonymous

Anonymous Tue 09 Jun 2020 17:34:57 No.11778756 Report

Quoted By: >>11779702

>>11774767
We dont already have the algorithms. Most of these systems are extremely limited and not generalizable.

You train the network on one thing and it fails on everything else.

Anonymous

View Same Google iqdb SauceNAO 1521850702964.png, 627KiB, 936x1301

Anonymous Tue 09 Jun 2020 18:03:09 No.11778824 Report

Quoted By:

>>11773541
Read Iron Man comic books.

Anonymous

Anonymous Tue 09 Jun 2020 20:24:00 No.11779339 Report

Quoted By: >>11779727

>>11774767
We don't even know what intelligence is.

Every algorithm known today fails in the chinese room argument.

>>11773541
To start I suggest this: http://www.hutter1.net/ai/introref.htm#longread

Anonymous

Anonymous Tue 09 Jun 2020 21:34:39 No.11779554 Report

Quoted By:

>>11773588
the material basis to consciousness is what renders ASI impossible

Anonymous

Anonymous Tue 09 Jun 2020 21:52:11 No.11779601 Report

Quoted By:

>>11773541
2025? Is that the prediction here?

Anonymous

Anonymous Tue 09 Jun 2020 22:30:48 No.11779702 Report

Quoted By:

>>11778756
>You train the network on one thing and it fails on everything else.
Not really true. Case in point: GPT-3. The whole point of the GPT-3 paper is to show that these huge models somehow are great at things they weren’t trained to do. That said, they are indeed still trained on limited tasks, but to train systems on “everything”, we need more computing power.

Anonymous

Anonymous Tue 09 Jun 2020 22:37:56 No.11779727 Report

Quoted By: >>11780420

>>11779339
>To start I suggest this: http://www.hutter1.net/ai/introref.htm#longread
That’s actually a great resource, but it contradicts your claim that
>We don't even know what intelligence is.

>Every algorithm known today fails in the chinese room argument.
So does every human.

DoctorGreen !DRgReeNusk

DoctorGreen !DRgReeNusk Tue 09 Jun 2020 22:55:48 No.11779794 Report

Quoted By:

>>11773541
biology

Anonymous

Anonymous Wed 10 Jun 2020 02:07:13 No.11780420 Report

Quoted By:

>>11779727
>but it contradicts your claim that
I mean, this link works for ANI, to AGI we first need to understand what intelligence is, the ANI study is an initial step to archive this.

>So does every human.
The CRA states that the computer will just operate the input into an output by computer instructions, the human we don't know if it operates an input into an output, we don't know how a human comprehend language.

Anonymous

View Same Google iqdb SauceNAO biggggs.png, 140KiB, 293x430

Anonymous Wed 10 Jun 2020 03:25:38 No.11780607 Report

Quoted By:

>>11773588
what is quantum consciousness?

Anonymous

Anonymous Wed 10 Jun 2020 03:27:46 No.11780613 Report

Quoted By: >>11780620

>>11774414
>Physics
Why?

Anonymous

Anonymous Wed 10 Jun 2020 03:29:16 No.11780620 Report

Quoted By: >>11781273

>>11780613
they don’t understand physics and think quantum computing and AI will synergize to bootstrap the singularity they’ve been sliding away from for over a decade now. CS niggers are legitimately disgusting retard dilettantes.

Anonymous

Anonymous Wed 10 Jun 2020 07:20:47 No.11781273 Report

Quoted By:

>>11780620
You think it won't? Can you provide coherent reasoning why?

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches