/sci/ - Science & Math » Thread #11883453

44KiB, 800x353, Picture01.png

View Same Google iqdb SauceNAO

Anonymous Thu 09 Jul 10:44:14 2020 No.11883453 View Reply Original Report

Quoted By: >>11883457 >>11883949 >>11883970 >>11884011 >>11884013 >>11884377 >>11884384 >>11884776 >>11885022 >>11885053 >>11886436 >>11886454 >>11888916

When will it happen?

Anonymous

Anonymous Thu 09 Jul 2020 10:47:23 No.11883457 Report

Quoted By:

>>11883453
AGI? I predict 60 - 90 years from now.

Anonymous

Anonymous Thu 09 Jul 2020 15:53:14 No.11883949 Report

Quoted By:

>>11883453
mid 2023

Anonymous

Anonymous Thu 09 Jul 2020 16:01:43 No.11883970 Report

Quoted By: >>11883985 >>11884029

>>11883453
Matter of fact is that Ai is not superior in intelligence and it never will be. The superiority is an result of his massive ability to remember.
But the learning curve is beyond shit, if yo want some one learn sth. very fast i swear you you will prefer an mongoloid ova any ai.

Anonymous

Anonymous Thu 09 Jul 2020 16:08:50 No.11883985 Report

Quoted By: >>11884356

>>11883970
I tough about this experiments with octopuses, to determent there intelligence. I am now 100% certain they are many magnitudes better physicists as we are.

How can we integrate in our society? Teach theme to speak? Handle an pc and profit from there unique neuronal structure?
If this animal can solve mechanical problems, 100 times faster as we it will be at this factor a better physicist. How do we teach it to talk and convince it to work on our problems?

Anonymous

Anonymous Thu 09 Jul 2020 16:13:29 No.11883992 Report

Quoted By:

Lmao, when there is an actual revolution and AI stops being statistical inference irt.

Anonymous

View Same Google iqdb SauceNAO x go home.png, 8KiB, 1137x828

Anonymous Thu 09 Jul 2020 16:25:25 No.11884011 Report

Quoted By:

>>11883453
never ever

Anonymous

Anonymous Thu 09 Jul 2020 16:26:43 No.11884013 Report

Quoted By:

>>11883453
3000 AD

Anonymous

Anonymous Thu 09 Jul 2020 16:32:43 No.11884029 Report

Quoted By: >>11884040 >>11884052 >>11884241 >>11884254 >>11884800

>>11883970
OpenAI learned to play Dota2 at professional level in two weeks. Alphazero learned to play chess beyond human ability in 24 hours.

Anonymous

Anonymous Thu 09 Jul 2020 16:36:38 No.11884040 Report

Quoted By: >>11884082

>>11884029
The Dota2 one isn't a great example as the test maps and characters were very restricted. Also it basically won by having incredibly fast reaction times no human could ever achieve. That's not intelligence, that's gigahertz.

Anonymous

Anonymous Thu 09 Jul 2020 16:39:15 No.11884050 Report

Quoted By: >>11884481

The complexity of problems grows exponentially faster than the growth of intelligence and computation
AI won't be able to help us solve most problems. It won't make as drastic a difference to society and the human condition as many people hypothesize.

Anonymous

Anonymous Thu 09 Jul 2020 16:39:23 No.11884052 Report

Quoted By:

>>11884029
yeah AI doing well in the confines of a closed system with clearly delineated search space - really comparable to intelligence in the real world with nigh infinite dimensionality. wow AGI is coming right soon!!

Anonymous

Anonymous Thu 09 Jul 2020 16:50:25 No.11884082 Report

Quoted By: >>11884792

>>11884040
>The Dota2 one isn't a great example as the test maps and characters were very restricted. Also it basically won by having incredibly fast reaction times no human could ever achieve. That's not intelligence, that's gigahertz.
Not true, bots had 0.2s reaction time, similar to pro players. Reaction time didn't have effect. And no matter how you put it, Dota 2 is extremely complex game that requires strategic forward thinking and team collaboration. AIs are at early stage now, but already beating humans in such complex games. First chess, then go, then poker and then dota2. You see where it's going. Soon they will beat us in the game of life.

Anonymous

Anonymous Thu 09 Jul 2020 17:04:12 No.11884118 Report

Quoted By:

How many games does an "AI" have to play before it can be good? Millions of super-sped up ones? Does it learn as much from a single game as a human does (or more)?

I think the main hurdle will be for the AI to participate in a conversation and invent new approaches for dealing with problems, and communicating them to humans.

Anonymous

Anonymous Thu 09 Jul 2020 17:51:31 No.11884241 Report

Quoted By: >>11888178

>>11884029
The Ai needed 4 million tries to get on an human level. Additionally an Ai cant fail due to physical limitations, The Ai cant have a uneven mouse pad. With an instant mind transfer humans would be much fast. I know AI much of the AIs power is reaction time not intelligence.

4 Million trys!

Also the vector angels in real life are infinite, in an game the are limited, making it much more izzy.

Show me one AI who is smarter, not faster.

Anonymous

Anonymous Thu 09 Jul 2020 17:55:42 No.11884254 Report

Quoted By:

>>11884029
Alphazero played more grandmaster level chess match with itself than there have been in entire human history. Anon is not entirely wrong about ability to remember. For some of the most advanced methods it does hinge on tremendous computational power.
Although we're currently seeing more "lean" approaches emerging.

Anonymous

Anonymous Thu 09 Jul 2020 18:15:10 No.11884310 Report

Quoted By: >>11889046 >>11889288

"Computation" and "information" aren't real, they are human words to describe abstractions.
The physical particles that make up our brains and such have completely different wave-functions and connection capabilities than the silicon chips used in modern machines. The only reason some people think that a computer can think like a person is because of the idea of the universal Turing machine. But computation and information aren't real. The philosophical idea of a turing machine, as a tape with a stamp and a library of symbols etc. is simply not at all how our brains and minds ACTUALLY WORK. their process is a physical process that can only be replicated on equivalent physical material (i.e. carbon based neurons arranged in this way).
Things like the universal approximation theorem and the church turing thesis and such are entirely irrelevant with respect to the physics of intelligence.

Anonymous

View Same Google iqdb SauceNAO kvkgtma7od911.jpg, 94KiB, 672x737

Anonymous Thu 09 Jul 2020 18:25:58 No.11884347 Report

Quoted By:

i really don't see how "very smart" AI could fix or solve any of the problems we have. for starters let's throw out all A.I that imitates human behavior because it's consumer gimmicky garbage and let's focus on A.I that deals with data. first of all AI's observation is the same or even less than ours meaning that its conclusions won't be much different than ours.
second of all 99% of current problems aren't "data too complex for humans to interpret". it's just problems that can't be solved or are to be solved. A.I won't desalinate water, A.I won't stop entropy, A.I won't solve gravity, etc.
i think A.I will be a slight boost to human progress and advancement but it will never be what nick land told you it'd be.

Anonymous

Anonymous Thu 09 Jul 2020 18:30:08 No.11884356 Report

Quoted By:

>>11883985
So put an octopiss brain in your self drivig car faggot. Oh no you won't hahaha. Retard

Anonymous

Anonymous Thu 09 Jul 2020 18:35:33 No.11884377 Report

Quoted By:

>>11883453
Right now. You're being made fun of by google captcha when you try to make a post. You just gonna sit there and take it like a little bitch?

Anonymous

Anonymous Thu 09 Jul 2020 18:38:34 No.11884384 Report

Quoted By: >>11884409 >>11888158

>>11883453
Stop buying sci fi fairytales. We can never recreate the intricacy of the human brain in a machine let alone give rise to any sort of super AI.

Anonymous

Anonymous Thu 09 Jul 2020 18:39:09 No.11884386 Report

Quoted By:

tuesday

Anonymous

Anonymous Thu 09 Jul 2020 18:49:13 No.11884409 Report

Quoted By: >>11884425 >>11884701

>>11884384
>We can never recreate the intricacy of the human brain in a machine
It would be easy. You're not speaking from a scientific basis, but an emotional one.

Anonymous

Anonymous Thu 09 Jul 2020 18:56:11 No.11884425 Report

Quoted By: >>11884517

>>11884409
And you know that how?

Anonymous

Anonymous Thu 09 Jul 2020 19:18:27 No.11884481 Report

Quoted By:

>>11884050
The big differences in society will come cause we'll have something that's (at least) as smart as a human and that you can create in batches with CTRL-C CTRL-V

It doesn't matter if it won't be able to magically solve hard problems, change will be dramatic anyhow

Anonymous

Anonymous Thu 09 Jul 2020 19:33:12 No.11884517 Report

Quoted By: >>11884610 >>11884701 >>11884764

>>11884425
Because we already understand the human brains structure, and we know that the essential elements can be almost directly replicated using digital circuitry. If you paid any attention at all to developments in neuroscience instead of hanging out on /lit/ moping about philosophy, you would know this.

Anonymous

Anonymous Thu 09 Jul 2020 20:18:44 No.11884610 Report

Quoted By:

>>11884517
We know jack shit about the brain, neuroscience is in its infancy, and even if we knew everything we needed to know we wouldn't be able to make a machine that's capable of harnessing all that.

Go outside.

Anonymous

Anonymous Thu 09 Jul 2020 20:46:32 No.11884701 Report

Quoted By:

>>11884409
False
>>11884517
>Because we already understand the human brains structure, and we know that the essential elements can be almost directly replicated using digital circuitry.
False.

Anonymous

View Same Google iqdb SauceNAO B26F620F-277B-4CE7-9874-C10A3D9B (...).jpg, 52KiB, 500x500

Anonymous Thu 09 Jul 2020 21:05:19 No.11884764 Report

Quoted By:

>>11884517
>implying 3D wetware can be 1-to-1 duplicated in digital space
>implying (((neural nets))) are even close to a faithful representation of a biological neuron

Anonymous

View Same Google iqdb SauceNAO 1585167459390.jpg, 143KiB, 821x1024

Anonymous Thu 09 Jul 2020 21:09:28 No.11884776 Report

Quoted By:

>>11883453
Depends on how you define "human intelligence" For example most women fail the Turing test.

Anonymous

Anonymous Thu 09 Jul 2020 21:14:22 No.11884792 Report

Quoted By:

>>11884082
i think after humans lost to the bot they learned strategies the bot uses and can easily beat it now. the problem is that the bot doesn't adopt like that, it just uses its knowledge from training and doesn't learn from there, whereas a human can learn from 1 or 2 games and drastically modify their play style.

Anonymous

View Same Google iqdb SauceNAO BAF4EAAF-8D92-45EF-AF28-9D010B00 (...).jpg, 157KiB, 1000x667

Anonymous Thu 09 Jul 2020 21:16:37 No.11884800 Report

Quoted By:

>>11884029
>my neural net is intellig-
*blocks your path*

Anonymous

View Same Google iqdb SauceNAO 000021.png, 12KiB, 700x432

Anonymous Thu 09 Jul 2020 22:19:48 No.11885022 Report

Quoted By:

>>11883453
I too can make up graphs

Anonymous

Anonymous Thu 09 Jul 2020 22:28:22 No.11885053 Report

Quoted By:

>>11883453
imagine how good the memes created by artificial super intelligence will be

Anonymous

Anonymous Fri 10 Jul 2020 08:49:05 No.11886436 Report

Quoted By:

>>11883453
>Human intelligence is one long straight line
Nice bait

Anonymous

Anonymous Fri 10 Jul 2020 08:53:56 No.11886444 Report

Quoted By:

100 trillion parameter models when?

Anonymous

Anonymous Fri 10 Jul 2020 09:03:19 No.11886454 Report

Quoted By: >>11886456

>>11883453
IT'S NOT GOING TO HAPPEN EVER. Only white people could do this and they have decided to dress nigger up in European clothing, give them drivers licenses and pretend they're people. Then they are going to sacrifice themselves to the BigJewMedia to prove they are not racist, literally the worst thing you can be according to the Jews.

Anonymous

Anonymous Fri 10 Jul 2020 09:04:46 No.11886456 Report

Quoted By:

>>11886454
Faggot cuck white people. What a bunch of easily manipulate autists.

Anonymous

Anonymous Fri 10 Jul 2020 21:17:16 No.11888158 Report

Quoted By: >>11889046

>>11884384
>We can never recreate the intricacy of the human brain in a machine let alone give rise to any sort of super AI.
Why not? If evolution could give rise to the human brain though nothing more than millions of years of trial and error I don't see why a sufficiently large machine learning algorithm could do the same if given enough time.

Anonymous

Anonymous Fri 10 Jul 2020 21:21:42 No.11888178 Report

Quoted By: >>11889046

>>11884241
>The Ai needed 4 million tries to get on an human level.
Life needed several million generations to get to a human level, your point?

Anonymous

Anonymous Sat 11 Jul 2020 02:55:14 No.11888916 Report

Quoted By: >>11889050

>>11883453
Already happened

Anonymous

Anonymous Sat 11 Jul 2020 04:10:26 No.11889046 Report

Quoted By:

>>11888158
>>11888178
see >>11884310

Anonymous

View Same Google iqdb SauceNAO 1590860310257.jpg, 229KiB, 527x762

Anonymous Sat 11 Jul 2020 04:11:11 No.11889050 Report

Quoted By:

>>11888916
Correct. As of right now these smarter and faster than human AI's are all in the money making business trading stocks/futures/currencies etc. The governments are using them behind the scenes for more nefarious purposes.

Anonymous

Anonymous Sat 11 Jul 2020 06:13:33 No.11889288 Report

Quoted By:

>>11884310
>The philosophical idea of a turing machine, as a tape with a stamp and a library of symbols etc. is simply not at all how our brains and minds ACTUALLY WORK
:c

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