EPTStarCraft II: DeepMind Demonstration: Jan 24 - Page 24
| Forum Index > SC2 General |
|
[F_]aths
Germany3947 Posts
| ||
|
siakam
Korea (South)11 Posts
No gg no skill  | ||
|
renaissanceMAN
United States1840 Posts
On January 25 2019 05:21 FFW_Rude wrote: That's called CounterStrike mentality. You are a noob or a cheater. No middleground good thing I'm fucking garbage at cs:go | ||
|
The_Red_Viper
19533 Posts
On January 25 2019 05:19 ZigguratOfUr wrote: Alphastar leaving the base with his army repeatedly and Mana's warp prism just going back in immediately doesn't speak too highly of its decision making ability. But I'm sure it'll get better. That's a very specific case though, at some point it will build a phoenix when something like this happens and then win the game and thus be reinforced to build a phoenix more often when something like this happens and thus learn. | ||
Nakajin
Canada8989 Posts
On January 25 2019 05:19 Poopi wrote: Yeah going back and forth to defend the prism, not making a phoenix and stuff was really bad. Sure but the oracle play was great, that 2 oracle in the main with a front pressure was nasty, would have won against 99% of the players. | ||
|
Zaros
United Kingdom3692 Posts
| ||
|
IshinShishi
Japan6156 Posts
On January 25 2019 05:22 The_Red_Viper wrote: That's a very specific case though, at some point it will build a phoenix when something like this happens and then win the game and thus be reinforced to build a phoenix more often when something like this happens and thus learn. yea, I think the other version faced this in it's 200 years of training, this one is just a lil pup | ||
|
[F_]aths
Germany3947 Posts
 | ||
|
Poopi
France12911 Posts
On January 25 2019 05:22 Nakajin wrote: Sure but the oracle play was great, that 2 oracle in the main with a front pressure was nasty, would have won against 99% of the players. I don't think you should judge how an AI like that really performs based on its strong showings (often due to inherent advantages), but on how robust it is to trickeries. | ||
|
_fool
Netherlands682 Posts
On January 24 2019 05:20 _fool wrote: I think we will be amazed. I was amazed | ||
|
Kazi25
Philippines236 Posts
| ||
|
Xamo
Spain886 Posts
On January 25 2019 05:25 Poopi wrote: I don't think you should judge how an AI like that really performs based on its strong showings (often due to inherent advantages), but on how robust it is to trickeries. The DeepMind team seems to agree, they specifically told us that they picked the "least exploitable agents" from their league. | ||
|
Xamo
Spain886 Posts
| ||
|
Lanthdoral
Germany7 Posts
| ||
|
Deleted User 3420
24492 Posts
| ||
|
RandomPlayer
Russian Federation400 Posts
| ||
|
Aiingel
243 Posts
| ||
|
FFW_Rude
France10201 Posts
"AlphaStarAi says : GIMME SERRAL OR I'M GONNA KILL CORDANA" made me laugh | ||
|
SPcrusader
Norway99 Posts
| ||
|
Nakajin
Canada8989 Posts
On January 25 2019 05:25 Poopi wrote: I don't think you should judge how an AI like that really performs based on its strong showings (often due to inherent advantages), but on how robust it is to trickeries. Well the strategt the IA came up with was both very good vs what Mana was doing and didn't depend on any inherent advantages and it dealt well enough with the adepts harras, sure clearly it didn't understand how to respond to immortal-WP but it still look like an IA who would be unbeatable in a few week, at least in pvp on this map. | ||
| ||
