[ExI] Superhuman Poker
William Flynn Wallace
foozler83 at gmail.com
Wed Jul 17 23:29:22 UTC 2019
> Can we learn what it learned?
Not really, the program can't tell us why it's so good at what it does for
the same reason a human genius can't tell us why he's so good at what he
does, neither of them knows.
John K Clark
It's right there in the code, isn't it? But can't someone tell what
changes have been made from the original code, which was written by
humans? I am way out of my depth here, but if AIs are going to learn
things for us, somehow we should be able to tell how it does what it does.
Maybe not now but in the future? Otherwise the AIs have all the secrets!
On Wed, Jul 17, 2019 at 6:09 PM John Clark <johnkclark at gmail.com> wrote:
> On Tue, Jul 16, 2019 at 10:40 PM William Flynn Wallace <
> foozler83 at gmail.com> wrote:
> > Impressive. Now - the AI programmed itself somewhat after a lot of
>> play, I assume.
> Yes, it played against itself millions of times during the training period
> but required no input from human players. The training period needed 512 GB
> of memory and 8 days on a 64-core Intel Xeon E5-8860 server for a total of
> 12,288 CPU core hours. If you bought that much processing power on cloud
> computing it would only cost you $144. After the training period was
> complete and a winning strategy was found on how to play Poker at a
> superhuman level then during a real game in real time against real
> opponents the program only needed 128 GB of memory.
>> > Can we learn what it learned?
> Not really, the program can't tell us why it's so good at what it does for
> the same reason a human genius can't tell us why he's so good at what he
> does, neither of them knows.
> John K Clark
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat