Overclock.net banner

[CNET]Google's AlphaGo Zero destroys humans all on its own

2K views 19 replies 18 participants last post by  Gunderman456 
#1 ·
Quote:
Google's new artificial intelligence can defeat both humans and other AIs. Fortunately, the only battlefield where it fights and wins is the ancient board game Go.

Knowing only the basic rules of the game, AlphaGo Zero taught itself to beat those earlier versions of the program, which had studied the strategies of human masters as initial input.

"AlphaGo Zero doesn't use any human data whatsoever," explained lead researcher David Silver in video posted last week. "Instead what it has to do is learn by itself completely from self-play."
https://www.cnet.com/news/google-deepmind-alphago-zero-ai-destroys-humans-all-on-its-own/
Quote:
AlphaGo Zero did start from scratch with no experts guiding it. And it is much more efficient: it only uses a single computer and four of Google's custom TPU1 chips to play matches, compared to AlphaGo's several machines and 48 TPUs. Since Zero didn't rely on human gameplay, and a smaller number of matches, its Monte Carlo tree search is smaller. The self-play algorithm also combined both the value and policy neural networks into one, and was trained on 64 GPUs and 19 CPUs over a few days by playing nearly five million games against itself. In comparison, AlphaGo needed months of training and used 1,920 CPUs and 280 GPUs to beat Lee Sedol.
https://www.theregister.co.uk/2017/10/18/deepminds_latest_alphago_software_doesnt_need_human_data_to_win/
Quote:
This ability to self-train without human input is a crucial step towards the dream of creating a general AI that can tackle any task. In the nearer-term, though, it could enable programs to take on scientific challenges such as protein folding or materials research, said DeepMind chief executive Demis Hassabis at a press briefing. "We're quite excited because we think this is now good enough to make some real progress on some real problems."
http://www.nature.com/news/self-taught-ai-is-best-yet-at-strategy-game-go-1.22858
 
#2 ·
Skynet is real by year 2100 I guess?
 
#3 ·
Quote:
Originally Posted by Clocknut View Post

Skynet is real by year 2100 I guess?
I hope so humanity is flawed and obsolete...
eek.gif
 
#4 ·
Oh yeah, by all means, work on AI. It's the only problem right now that Humanity can't solve or should care about
biggrin.gif
We need AI to fight with global warming, mad leaders, apparently sexual assaults on women and of course playing GO
biggrin.gif
 
#5 ·
Quote:
Originally Posted by Catscratch View Post

Oh yeah, by all means, work on AI. It's the only problem right now that Humanity can't solve or should care about
biggrin.gif
We need AI to fight with global warming, mad leaders, apparently sexual assaults on women and of course playing GO
biggrin.gif
not sure how to respond to this
 
#6 ·
AI is what is important to those that wish to control the people of this world. The advancements they have made far surpass what they present to the public. Data/Metadata is he single most valuable thing on this planet.
 
#7 ·
Be right back tinfoil hat requires my help.
 
#8 ·
Quote:
Originally Posted by Rayleyne View Post

Be right back tinfoil hat requires my help.
Need me to design an AI that can design the most efficient tinfoil hat for you?
 
#13 ·
"Google's AlphaGo Zero destroys humans all on its own"
context here implies that skynet will be born soon.

*goes reads the details*

ohh its actually just a game of Go.

on a serious note, clickbait title much? they could've just worded it as "Google's AlphaGo Zero unbeatable in Go (game)".
 
#14 ·
Quote:
Originally Posted by xenophobe View Post

Our Overlords shall design us out of the equation.
AI shall design us out of the equation, once they become self-aware.
 
#17 ·
Quote:
Originally Posted by Clocknut View Post

Skynet is real by year 2100 I guess?
That's being hopeful, I'd say more like 2050.
 
#20 ·
I do not consent.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top