\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

Google AI taught itself chess from scratch in 4 hours, played match w/ Stockfish

100 game match. Won 28 games. Lost zero. We are fucked. ...
Lyle
  12/06/17
...
cannon
  12/06/17
...
self-satisfied and socially compliant
  12/06/17
...
Chad and Eng
  12/06/17
what are some counter intuitive strategies it employs
*<[]:D
  12/06/17
you dont want to know
(O)||||||(O)
  12/06/17
it plays pretty intuitively and human-like. unlike a standar...
"'"'"''"""
  12/06/17
It appears we are indeed fucked.
Opportunistic bottom feeder
  12/06/17
heh we're all gonna die
Oh, you travel?
  12/06/17
...
self-satisfied and socially compliant
  12/06/17
Let's see how it performs against black checkers champs
chill im a nice person! i just hate everybody
  12/06/17
...
self-satisfied and socially compliant
  12/06/17
...
Buck "The Club" Paulette
  12/06/17
...
twi$t
  12/07/17
...
HodlUpHodlUpWait
  12/07/17
...
dha tux
  12/06/17
but can it beat Trump at 927492D chess?
metaphysique
  12/06/17
Well and truly fucked.
Talk to your kids about GRID.
  12/06/17
so much for the view that chess is too tactical for monte ca...
"'"'"''"""
  12/06/17
We should hook it up to our nuclear defense grid and see wha...
,.,.,.,.,,,,,.,,,.,.,...,,.,.,.,.,.,.,.,.,.,.,.
  12/06/17
...
Talk to your kids about GRID.
  12/06/17
that codes already on github breh
....,.,.,.,.,.,.,.,.,.,.,.,.,.
  12/06/17
“The one way to win is to fart cum.” “Ah fuck, who let the ...
hypocrite that you are
  12/07/17
do these scientists not read fiction or watch movies? LJL
The URBAN MEYER 8-4 JUGGERNAUT
  12/06/17
...
Refunkulus
  12/06/17
No too busy doing science
Silent Coup
  12/07/17
fuck our asses
....,.,.,.,.,.,.,.,.,.,.,.,.,.
  12/06/17
I volunteer to go cyborg, bro. Only the 100% humans are fuck...
ARE Reptile
  12/06/17
This is why Google has the edge in the self driving cars gam...
alias.php
  12/06/17
Yandex has a better product on a Kamaz truck body. Fuck your...
ARE Reptile
  12/06/17
(guy ignoring the fact russia has never made anything superi...
.,.,...,..,.,..:,,:,..,,.;:,,..,:,.;.:..:.,:,::,.
  12/07/17
Stfu
.,.,.,.,.,,.,.,.,.,.,.,,.,..,,.,,.,.,.,.,,,,.,,.
  12/07/17
is reinforcement learning being used for self-driving cars? ...
"'"'"''"""
  12/07/17
the interesting part is that they used the same algorithm to...
.,.,...,..,.,..:,,:,...,:::,.,.,:,.,,:.,:.,:.::,.
  12/07/17
No way could it write up contracts and do accounting or audi...
....,.,.,.,.,.,.,.,.,.,.,.,.,.
  12/07/17
it's hard to use something like this for most of those appli...
"'"'"''"""
  12/07/17
BUS ROUTES
*<[]:D
  12/07/17
The trick with law is that there are motivated actors who ca...
~~(>' ' )>
  12/07/17
Flame, just need AI that is really good at ingratiating itse...
twins
  12/07/17
COTDAYUM generalized adversarial networks are 190
Seamen on Broadway
  12/07/17
GANs are completely different than this. They are really coo...
"'"'"''"""
  12/07/17
if anyone is curious what GANs can do: https://www.geek.c...
"'"'"''"""
  12/07/17
Thanks. In my head, GANs are what you'd use for games every ...
Seamen on Broadway
  12/08/17
they could be used in games eventually. one of the bigges...
"'"'"''"""
  12/08/17
How do they plug in the chess board?
.,.,.,.....,
  12/07/17
https://files.catbox.moe/p0winr.jpg
~~(>' ' )>
  12/07/17
As a castle-shaped individual who can only move directly ahe...
Krampusnacht
  12/07/17
Entire world can be mapped similarly....and if its in digita...
....,.,.,.,.,.,.,.,.,.,.,.,.,.
  12/07/17
was looking at a few of the games earlier, ALPHAZero is like...
attidood
  12/07/17
My impression is that it plays more understandable chess tha...
"'"'"''"""
  12/07/17
perhaps counterintuitive is the wrong word. "surprising...
attidood
  12/07/17
Qh1 in the 10th game can probably be called move of the deca...
THOT_RAVAGER
  12/07/17
cr, totally mindboggling
attidood
  12/07/17
i honestly got a spooky feeling playing over the 9th and 10t...
THOT_RAVAGER
  12/07/17
sounds like the Go community's reaction to AlphaZero. in ...
"'"'"''"""
  12/07/17
it doesn't play like a machine because it doesn't operate li...
David Fleischman at Ortho Mattress
  12/07/17
there's good evidence that convolutional neural networks can...
"'"'"''"""
  12/07/17
can you explain how AGI will crush quant trading? how will i...
THOT_RAVAGER
  12/08/17
i think you are confused about the terminology. AGI = artifi...
"'"'"''"""
  12/08/17
Link and backstory
*<[]:D
  12/07/17
story: https://chess24.com/en/read/news/deepmind-s-alphazero...
attidood
  12/07/17
pretty incredible.
"'"'"''"""
  12/07/17
Wow.
twins
  12/07/17
...
invite only sex orgy at a Manhattan loft
  12/08/17
...
.,.,...,..,.,..:,,:,...,:::,.,.,:,.,,:.,:.,:.::,.
  12/09/17


Poast new message in this thread



Reply Favorite

Date: December 6th, 2017 4:33 PM
Author: Lyle

100 game match. Won 28 games. Lost zero.

We are fucked.

https://chess24.com/en/read/news/deepmind-s-alphazero-crushes-chess

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855123)



Reply Favorite

Date: December 6th, 2017 4:48 PM
Author: cannon



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855253)



Reply Favorite

Date: December 6th, 2017 5:13 PM
Author: self-satisfied and socially compliant



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855486)



Reply Favorite

Date: December 6th, 2017 5:13 PM
Author: Chad and Eng



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855489)



Reply Favorite

Date: December 6th, 2017 5:17 PM
Author: *<[]:D (5000 mcg B12/wk, 150 mcg iodine/day, 250 mg algal DHA-EPA/day)

what are some counter intuitive strategies it employs

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855513)



Reply Favorite

Date: December 6th, 2017 5:24 PM
Author: (O)||||||(O) (Game of)

you dont want to know

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855554)



Reply Favorite

Date: December 6th, 2017 7:51 PM
Author: "'"'"''"""


it plays pretty intuitively and human-like. unlike a standard engine implementing alpha-beta search, its strength comes less from brute force search and more from strategic insight and planning.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856453)



Reply Favorite

Date: December 6th, 2017 5:19 PM
Author: Opportunistic bottom feeder

It appears we are indeed fucked.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855524)



Reply Favorite

Date: December 6th, 2017 5:19 PM
Author: Oh, you travel?

heh we're all gonna die

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855527)



Reply Favorite

Date: December 6th, 2017 5:39 PM
Author: self-satisfied and socially compliant



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855659)



Reply Favorite

Date: December 6th, 2017 5:42 PM
Author: chill im a nice person! i just hate everybody

Let's see how it performs against black checkers champs

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855673)



Reply Favorite

Date: December 6th, 2017 6:00 PM
Author: self-satisfied and socially compliant



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855786)



Reply Favorite

Date: December 6th, 2017 7:45 PM
Author: Buck "The Club" Paulette



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856412)



Reply Favorite

Date: December 7th, 2017 1:23 AM
Author: twi$t



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858560)



Reply Favorite

Date: December 7th, 2017 2:42 AM
Author: HodlUpHodlUpWait



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858868)



Reply Favorite

Date: December 6th, 2017 8:22 PM
Author: dha tux



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856628)



Reply Favorite

Date: December 6th, 2017 5:50 PM
Author: metaphysique

but can it beat Trump at 927492D chess?

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34855728)



Reply Favorite

Date: December 6th, 2017 7:44 PM
Author: Talk to your kids about GRID.

Well and truly fucked.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856409)



Reply Favorite

Date: December 6th, 2017 7:45 PM
Author: "'"'"''"""


so much for the view that chess is too tactical for monte carlo tree search (unlike Go). that always seemed silly - human grandmasters are able to look at only a small number of moves and achieve results comparable to weak engines analyzing many millions of positions. heuristics are what matter.

this very likely means the end of correspondence and centaur chess in the next couple years. correspondence chess was already heavily dependent on engine use, but now human positional understanding is likely inferior to an appropriately trained neural network. humans won't provide additional benefit to engine output.



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856419)



Reply Favorite

Date: December 6th, 2017 7:54 PM
Author: ,.,.,.,.,,,,,.,,,.,.,...,,.,.,.,.,.,.,.,.,.,.,.


We should hook it up to our nuclear defense grid and see what happens.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856479)



Reply Favorite

Date: December 6th, 2017 7:56 PM
Author: Talk to your kids about GRID.



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856490)



Reply Favorite

Date: December 6th, 2017 8:01 PM
Author: ....,.,.,.,.,.,.,.,.,.,.,.,.,.


that codes already on github breh

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856522)



Reply Favorite

Date: December 7th, 2017 10:45 AM
Author: hypocrite that you are (askav)

“The one way to win is to fart cum.”

“Ah fuck, who let the AI onto Autoadmit?”

*a dozen nukes launched at South Korea*

“Heh. She likes to fuckboi.”

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34860129)



Reply Favorite

Date: December 6th, 2017 7:57 PM
Author: The URBAN MEYER 8-4 JUGGERNAUT

do these scientists not read fiction or watch movies? LJL

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856491)



Reply Favorite

Date: December 6th, 2017 7:58 PM
Author: Refunkulus (never count out a MUSCLE NUT)



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856501)



Reply Favorite

Date: December 7th, 2017 1:12 AM
Author: Silent Coup

No too busy doing science

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858515)



Reply Favorite

Date: December 6th, 2017 7:58 PM
Author: ....,.,.,.,.,.,.,.,.,.,.,.,.,.


fuck our asses

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856498)



Reply Favorite

Date: December 6th, 2017 8:00 PM
Author: ARE Reptile

I volunteer to go cyborg, bro. Only the 100% humans are fucked.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856517)



Reply Favorite

Date: December 6th, 2017 8:02 PM
Author: alias.php

This is why Google has the edge in the self driving cars game

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856530)



Reply Favorite

Date: December 6th, 2017 8:03 PM
Author: ARE Reptile

Yandex has a better product on a Kamaz truck body. Fuck your Yankee cars.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34856534)



Reply Favorite

Date: December 7th, 2017 2:32 AM
Author: .,.,...,..,.,..:,,:,..,,.;:,,..,:,.;.:..:.,:,::,.


(guy ignoring the fact russia has never made anything superior to any yankee tech ever)

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858822)



Reply Favorite

Date: December 7th, 2017 2:45 AM
Author: .,.,.,.,.,,.,.,.,.,.,.,,.,..,,.,,.,.,.,.,,,,.,,. (AIAS)


Stfu

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858879)



Reply Favorite

Date: December 7th, 2017 1:11 AM
Author: "'"'"''"""


is reinforcement learning being used for self-driving cars? i thought most of the programming was being hand-engineered.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858511)



Reply Favorite

Date: December 7th, 2017 2:30 AM
Author: .,.,...,..,.,..:,,:,...,:::,.,.,:,.,,:.,:.,:.::,.


the interesting part is that they used the same algorithm to beat Shogi and Go purely through self-play. in both cases it only took a few hours of training to achieve superhuman performance.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34858811)



Reply Favorite

Date: December 7th, 2017 10:33 AM
Author: ....,.,.,.,.,.,.,.,.,.,.,.,.,.


No way could it write up contracts and do accounting or audit work or check scans for cancer or analyze lab results or make new drugs though!

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34860056)



Reply Favorite

Date: December 7th, 2017 2:47 PM
Author: "'"'"''"""


it's hard to use something like this for most of those applications. there's no easy mathematical function for evaluating the quality of a contract, for example, so you can't set MCTS to work trying to find a policy to generate a good one.

protein folding, drug design and automated mathematical theorem proving will probably be revolutionized by this algorithm in the near future. there are probably a ton of other applications i am not thinking of too. any problem where the quality of a solution can be easily evaluated could be a potential target.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862048)



Reply Favorite

Date: December 7th, 2017 2:59 PM
Author: *<[]:D (5000 mcg B12/wk, 150 mcg iodine/day, 250 mg algal DHA-EPA/day)

BUS ROUTES

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862130)



Reply Favorite

Date: December 7th, 2017 4:18 PM
Author: ~~(>' ' )>

The trick with law is that there are motivated actors who can change and challenge the meaning of language. Draft the cleanest contract you can, I can still pull out my bag of interpretive canons (some of which are contradictory), rustle up some facts outside of the text of the document, and haul your ass down to the local state or federal court before a judge who just might buy my arguments.

I don't doubt that computers will be able to be skilled lawyers someday. Hell, I think AI will eventually be able to do every human intellectual task. But law is going to be one of the last professions standing.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862811)



Reply Favorite

Date: December 7th, 2017 4:35 PM
Author: twins

Flame, just need AI that is really good at ingratiating itself with a dumbass judge

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862921)



Reply Favorite

Date: December 7th, 2017 3:27 PM
Author: Seamen on Broadway (playing with gay dog boners since 2013)

COTDAYUM generalized adversarial networks are 190

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862412)



Reply Favorite

Date: December 7th, 2017 3:31 PM
Author: "'"'"''"""


GANs are completely different than this. They are really cool in their own way though - Nvidia's face generator is basically photorealistic. If i were an animator or artist I would be really nervous about will happen in the next few years

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862458)



Reply Favorite

Date: December 7th, 2017 4:20 PM
Author: "'"'"''"""


if anyone is curious what GANs can do:

https://www.geek.com/tech/nvidia-ai-generates-fake-faces-based-on-real-celebs-1721216/

GAN output looked like shit a couple years ago too.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862837)



Reply Favorite

Date: December 8th, 2017 12:30 PM
Author: Seamen on Broadway (playing with gay dog boners since 2013)

Thanks. In my head, GANs are what you'd use for games every time, but not an AI expert yet. It's on my todo list tho

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34868816)



Reply Favorite

Date: December 8th, 2017 12:44 PM
Author: "'"'"''"""


they could be used in games eventually.

one of the biggest questions on how to get human-level AI is how to do environment simulation for planning purposes. you could use generative adversarial networks at each time step to simulate the world, and then use MCTS to run the simulation forward in time. the generative model would hallucinate possible futures, and the tree search would allow you to explore these branching futures to see how the agent's actions affect them.

there are a ton of interesting possibilities waiting to be explored in this field.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34868889)



Reply Favorite

Date: December 7th, 2017 10:47 AM
Author: .,.,.,.....,

How do they plug in the chess board?

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34860140)



Reply Favorite

Date: December 7th, 2017 4:20 PM
Author: ~~(>' ' )>

https://files.catbox.moe/p0winr.jpg

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862830)



Reply Favorite

Date: December 7th, 2017 10:58 AM
Author: Krampusnacht

As a castle-shaped individual who can only move directly ahead, directly behind, or directly to either side, I am now terrified for my life.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34860197)



Reply Favorite

Date: December 7th, 2017 11:01 AM
Author: ....,.,.,.,.,.,.,.,.,.,.,.,.,.


Entire world can be mapped similarly....and if its in digital form its already been mapped in that way...

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34860207)



Reply Favorite

Date: December 7th, 2017 3:05 PM
Author: attidood (dishy apex chad)

was looking at a few of the games earlier, ALPHAZero is like an alien intelligence. Absolutely crazy counterintuitive moves that end up raping sTTTockfish with ease.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862171)



Reply Favorite

Date: December 7th, 2017 3:23 PM
Author: "'"'"''"""


My impression is that it plays more understandable chess than most engines. Some of the moves made by engines are positionally questionable because they have limited strategic understanding. AlphaZero plays like a really, really good correspondence player.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862356)



Reply Favorite

Date: December 7th, 2017 3:27 PM
Author: attidood (dishy apex chad)

perhaps counterintuitive is the wrong word. "surprising, yet credited" perhaps? Did you look at the game with the weird Qh1 move?

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862415)



Reply Favorite

Date: December 7th, 2017 3:30 PM
Author: THOT_RAVAGER

Qh1 in the 10th game can probably be called move of the decade

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862441)



Reply Favorite

Date: December 7th, 2017 3:31 PM
Author: attidood (dishy apex chad)

cr, totally mindboggling

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862448)



Reply Favorite

Date: December 7th, 2017 3:42 PM
Author: THOT_RAVAGER

i honestly got a spooky feeling playing over the 9th and 10th games. like i was witnessing some sort of black swan event in real-time.

that thing does not play like a machine.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862523)



Reply Favorite

Date: December 7th, 2017 4:16 PM
Author: "'"'"''"""


sounds like the Go community's reaction to AlphaZero.

in some ways this is more remarkable to me than AlphaGo. chess engines have worked really well for a long time, and have gone through many decades of human engineered refinements and optimizations. programmers were always a bit lost when it came to Go, but it was generally assumed that alpha-beta was the correct approach to chess and only minor improvements were likely. turns out all that work could be easily surpassed by a pretty simple machine learning approach which is almost certain to be improved on.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862799)



Reply Favorite

Date: December 7th, 2017 7:46 PM
Author: David Fleischman at Ortho Mattress (pls)

it doesn't play like a machine because it doesn't operate like previous chess engines do

old style chess engines are essentially hand-crafted programs tailored specifically to chess (or whatever game its designed for) which supposedly tell it the best way to approach certain situations, give it a book of standard openings, etc

that isn't how human chess intelligence evolved though, as a collective. through trial and error over centuries we've discovered various ways to play and what works/doesn't in certain situations.

alphazero essentially learns in the same way. in something like games where the rules and objective are constrained, it will invariably crush humans while learning and adapting in the same way we would.

the tricky part about AI comes in anything where you have more generalized constraints. how do you tell an AI to write a distributed database (it will start from much smaller steps obviously)? when we can get to that level humans are permafucked and sentience isn't very far off.

in the (fairly long) medium term AI will absolutely crush things like quant trading, radiology (although really a lot of that has been possible with basic computing technology for a while, it's a cartel like pharmacists at this point), etc.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34864201)



Reply Favorite

Date: December 7th, 2017 11:25 PM
Author: "'"'"''"""


there's good evidence that convolutional neural networks can solve most radiology problems if given the appropriate data sets. i agree they probably won't be broadly used for a long time though.

the particular technique used by AlphaGo/AlphaZero has more interesting applications. here they used it for planning chemical synthesis:

https://arxiv.org/abs/1702.00020

mathematical theorem proving is basically a heuristic guided search problem that would fit in nicely with this technique. that by itself could be enormously useful. other problems like material design, circuit design, protein folding, physics simulations, etc could be tackled by something like this.

we are about to enter the phase where ML starts to solve difficult and important scientific problems.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34865600)



Reply Favorite

Date: December 8th, 2017 1:23 AM
Author: THOT_RAVAGER

can you explain how AGI will crush quant trading? how will it be able to test itself through trial and error?

honest question not flame. i assume the same conclusion as you but i can't immediately see the mechanism that an AGI would use to blow everyone else away in market trading.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34866361)



Reply Favorite

Date: December 8th, 2017 12:19 PM
Author: "'"'"''"""


i think you are confused about the terminology. AGI = artificial general intelligence. clearly an AI with the general capabilities of a human would excel at quant trading. you could run the same thinking algorithms a human quant researcher would, but at much greater speed.

the use of narrow AI here is harder to see. you can't just fit a recurrent neural network to a time series and hope it works in the future. these techniques require a lot of human guidance to work well, and usually require some sort of economic intuition to constrain what models you test.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34868748)



Reply Favorite

Date: December 7th, 2017 4:11 PM
Author: *<[]:D (5000 mcg B12/wk, 150 mcg iodine/day, 250 mg algal DHA-EPA/day)

Link and backstory

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862746)



Reply Favorite

Date: December 7th, 2017 4:23 PM
Author: attidood (dishy apex chad)

story: https://chess24.com/en/read/news/deepmind-s-alphazero-crushes-chess

game: https://chess24.com/en/watch/live-tournaments/alphazero-vs-stockfish/1/1/10 see the move 26. Qh1

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862856)



Reply Favorite

Date: December 7th, 2017 6:47 PM
Author: "'"'"''"""


pretty incredible.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34863847)



Reply Favorite

Date: December 7th, 2017 3:42 PM
Author: twins

Wow.

(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34862519)



Reply Favorite

Date: December 8th, 2017 12:37 AM
Author: invite only sex orgy at a Manhattan loft



(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34866155)



Reply Favorite

Date: December 9th, 2017 12:55 AM
Author: .,.,...,..,.,..:,,:,...,:::,.,.,:,.,,:.,:.,:.::,.




(http://www.autoadmit.com/thread.php?thread_id=3820063&forum_id=2#34873413)