Author Topic: Alpha Go Zero  (Read 1342 times)

0 Members and 1 Guest are viewing this topic.

Offline Luis Dias

  • 211
Stronger than everything else, this new version learnt its own gameplay from first principles, rather than being trained by human example and human guidance.

Think on that for a second. This means that you could train it to play any other game as well.

This breakthrough is less of a wowzer than last year's match with Lee Sedol, but this is getting to Skynet levels of crazy.


 
Between this & openAI's super DotA bot this summer, it's been an astonishing year for machine learning, especially with regards to general purpose AIs.


 

Offline The E

  • He's Ebeneezer Goode
  • 213
  • Nothing personal, just tech support.
    • Steam
    • Twitter
What I found most interesting about the AlphaGo writeups were the Go experts looking at its plays and saying "That's a non-human move". I wonder what they would say about this versions' play style.
If I'm just aching this can't go on
I came from chasing dreams to feel alone
There must be changes, miss to feel strong
I really need lifе to touch me
--Evergrey, Where August Mourns

 

Offline Luis Dias

  • 211
Quote
AlphaGo Zero discovered a remarkable level of Go knowledge during its self-play training process. This included not only fundamental elements of human Go knowledge, but also non-standard strategies beyond the scope of traditional Go knowledge.

Figure 5 shows a timeline indicating when professional joseki (corner sequences) were discovered (Fig. 5a and Extended Data Fig. 2); ultimately AlphaGo Zero preferred new joseki variants that were previously unknown (Fig. 5b and Extended Data Fig. 3). Figure 5c shows several fast self-play games played at different stages of training (see Supplementary Information). Tournament length games played at regular intervals throughout training are shown in Extended Data Fig. 4 and in the Supplementary Information. AlphaGo Zero rapidly progressed from entirely random moves towards a sophisticated understanding of Go concepts, including fuseki (opening), tesuji (tactics), life-and-death, ko (repeated board situations), yose (endgame), capturing races, sente (initiative), shape, influence and territory, all discovered from first principles. Surprisingly, shicho (‘ladder’ capture sequences that may span the whole board)—one of the first elements of Go knowledge learned by humans—were only understood by AlphaGo Zero much later in training.

https://www.nature.com/nature/journal/v550/n7676/full/nature24270.html

 

Offline Bobboau

  • Just a MODern kinda guy
    Just MODerately cool
    And MODest too
  • 213
just found this article
https://www.wired.com/2016/03/two-moves-alphago-lee-sedol-redefined-future/
it's from last years match but it's a compelling read.
Bobboau, bringing you products that work... in theory
learn to use PCS
creator of the ProXimus Procedural Texture and Effect Generator
My latest build of PCS2, get it while it's hot!
PCS 2.0.3


DEUTERONOMY 22:11
Thou shalt not wear a garment of diverse sorts, [as] of woollen and linen together