d2jsp
Log InRegister
d2jsp Forums > Other Games > Other Games > Starcraft I & II > Alphastar Starcract Ai Destroys Human > Rip Humans.
Prev123Next
Add Reply New Topic New Poll
Member
Posts: 25,903
Joined: Mar 11 2016
Gold: 17,609.00
Feb 11 2019 11:42pm
Quote (Cpt_Ghost @ 11 Feb 2019 16:22)
the only reason for alphastar being hard to beat is that he does not make mistakes and its micro and macro is top notch
yea, it's not hard to come across AI's capable of that, the only difference at the moment is that alphastar is self learning on top of that, which in itself is a feat for sure. But the reason i'm not impressed is that its just not impressive to beat a human under these conditions because it isn't even close to a fair fight, so it is false to advertise alphastar with "beating pros in starcraft"


welcome to real life, friend.

You don't appreciate the ability of this AI if you actually think any schmuck that can code can make a bot this intelligent at a task such as this.

Blizzard couldn't make a bot this good with 100 years and an instruction manual.

We must learn from it,

Like chess.

Its funny. When computers got better than humans at chess we were relieved that comps above our elo play the same opening moves.

Impressive for humans sake that 4gate was the good choice.

For example in go, The top go champion in the world played the go bot. Go bot played moves humans all think are bad. Go bot won like 20 games to none wrecked dominated no chance for human. Go bot opening move in what humans know as theory and translated to starcraft meta would be the equivalent of taking all your workers and running to natural to mine at start of game, didn't make sense in human meta. But computer DOMINATED top Go champion, so Humans suck at GO.

Same thing as above happened with Shogi, aka Japanese chess.

It seems that humans aren't that bad at toss considering it used same strats humans do. AT THIS POINT I should add.

This post was edited by Phaggot on Feb 11 2019 11:49pm
Member
Posts: 2,901
Joined: Jun 25 2017
Gold: 17,767.00
Feb 12 2019 02:08am
Quote (Cpt_Ghost @ 11 Feb 2019 23:22)
the only reason for alphastar being hard to beat is that he does not make mistakes and its micro and macro is top notch
yea, it's not hard to come across AI's capable of that, the only difference at the moment is that alphastar is self learning on top of that, which in itself is a feat for sure. But the reason i'm not impressed is that its just not impressive to beat a human under these conditions because it isn't even close to a fair fight, so it is false to advertise alphastar with "beating pros in starcraft"


I think you are missing the point of what AI is

The great thing here, is that the perfect micro/macro isn't hardcoded into Alphastar, but something it learned itself. can you point me to any other AI that has these capabilities and learned it based on basic agendas?

their advertisement may be misleading, I dont really know or care - but advertisements are usually misleading and fights are rarely fair, I dont think these really matter here
Member
Posts: 59,349
Joined: Feb 15 2009
Gold: 0.00
Feb 12 2019 10:27am
Quote (Phaggot @ 12 Feb 2019 07:42)
welcome to real life, friend.

You don't appreciate the ability of this AI if you actually think any schmuck that can code can make a bot this intelligent at a task such as this.

Blizzard couldn't make a bot this good with 100 years and an instruction manual.

We must learn from it,

Like chess.

Its funny. When computers got better than humans at chess we were relieved that comps above our elo play the same opening moves.

Impressive for humans sake that 4gate was the good choice.

For example in go, The top go champion in the world played the go bot. Go bot played moves humans all think are bad. Go bot won like 20 games to none wrecked dominated no chance for human. Go bot opening move in what humans know as theory and translated to starcraft meta would be the equivalent of taking all your workers and running to natural to mine at start of game, didn't make sense in human meta. But computer DOMINATED top Go champion, so Humans suck at GO.

Same thing as above happened with Shogi, aka Japanese chess.

It seems that humans aren't that bad at toss considering it used same strats humans do. AT THIS POINT I should add.


yes he is like that because it is not limited to the same input as humans are
in chess or any other game the input does not make a difference
in starcraft it does very much
at this point the AI does not have to deal with a whole layer of decision making that has a huge effect on the human players eg how to spend your attention span, your actions etc
without that it was pointless to move forward from playing chess with bots because it's doing the same thing it did in those games.
the dev team specifically chose starcraft because it was a different and challenging enviroment for the reasons above + that it was not a full information game like chess where you can see very movement on the map

This post was edited by Cpt_Ghost on Feb 12 2019 10:28am
Member
Posts: 25,903
Joined: Mar 11 2016
Gold: 17,609.00
Feb 12 2019 10:35am
Quote (Cpt_Ghost @ 12 Feb 2019 11:27)
yes he is like that because it is not limited to the same input as humans are
in chess or any other game the input does not make a difference
in starcraft it does very much
at this point the AI does not have to deal with a whole layer of decision making that has a huge effect on the human players eg how to spend your attention span, your actions etc
without that it was pointless to move forward from playing chess with bots because it's doing the same thing it did in those games.
the dev team specifically chose starcraft because it was a different and challenging enviroment for the reasons above + that it was not a full information game like chess where you can see very movement on the map


You are complaining about the inherent fact of being a computer I don't understand your argument that is how life works.
Member
Posts: 59,349
Joined: Feb 15 2009
Gold: 0.00
Feb 12 2019 10:37am
Quote (Phaggot @ 12 Feb 2019 18:35)
You are complaining about the inherent fact of being a computer I don't understand your argument that is how life works.


I'm complaining about the fact that the absence of these limitations highly reduces the effectiveness of the research in my opinion
I might not be an AI research expert, but i know what makes this game hard and how it works and they are certainly not using it to it's full potential, even though they specifically chose it because of that said potential.
meh.
Member
Posts: 25,903
Joined: Mar 11 2016
Gold: 17,609.00
Feb 12 2019 11:16am
Quote (Cpt_Ghost @ 12 Feb 2019 11:37)
I'm complaining about the fact that the absence of these limitations highly reduces the effectiveness of the research in my opinion
I might not be an AI research expert, but i know what makes this game hard and how it works and they are certainly not using it to it's full potential, even though they specifically chose it because of that said potential.
meh.


Im pretty sure the stalker micro and constant micro during surround and perfect blinking is the fullest potential we have ever seen.

Everything else is just silly to say, obviously. Let the computer play itself for 1000 years maybe it learns 1 zealot 6 probe rushes to be perfect meta......

And your other argument is "they dont limit/dumb down the AI enough, this is stupid" which is a ridiculous argument. The whole point of this is to learn and improve based upon computer algorithms finding new vertexes and computing new efficient strats. Dumbing down the AI idk.

And you can't say its impossible for humans to do that, now that we see it, we know what to practice, etc etc etc

This is all Chess/Shogi/Dota/Go all over again
Member
Posts: 59,349
Joined: Feb 15 2009
Gold: 0.00
Feb 12 2019 11:18am
Quote (Phaggot @ 12 Feb 2019 19:16)
Im pretty sure the stalker micro and constant micro during surround and perfect blinking is the fullest potential we have ever seen.

Everything else is just silly to say, obviously. Let the computer play itself for 1000 years maybe it learns 1 zealot 6 probe rushes to be perfect meta......

And your other argument is "they dont limit/dumb down the AI enough, this is stupid" which is a ridiculous argument. The whole point of this is to learn and improve based upon computer algorithms finding new vertexes and computing new efficient strats. Dumbing down the AI idk.

And you can't say its impossible for humans to do that, now that we see it, we know what to practice, etc etc etc

This is all Chess/Shogi/Dota/Go all over again


that is the problem
i can see you do not understand a word i'm saying, so i'm not going to respond any further.
Member
Posts: 25,903
Joined: Mar 11 2016
Gold: 17,609.00
Feb 12 2019 11:44am
Quote (Cpt_Ghost @ 12 Feb 2019 12:18)
that is the problem
i can see you do not understand a word i'm saying, so i'm not going to respond any further.



Again, in your mind, to make it fair, we would need to limit the AI in terms of calcualting time loss for mouse movements and micro seconds lost with reaction time and button pressing, not to mention the clicking on minimap/unit control.

That isn't the point, that also is stupid. It isn't about whether or not we can make an AI that has the same detriments as humans as some sort of existential crisis solution.

That is an inherent property of being an AI playing a computer game.

This post was edited by Phaggot on Feb 12 2019 11:44am
Member
Posts: 59,349
Joined: Feb 15 2009
Gold: 0.00
Feb 12 2019 11:54am
Quote (Phaggot @ 12 Feb 2019 19:44)
Again, in your mind, to make it fair, we would need to limit the AI in terms of calcualting time loss for mouse movements and micro seconds lost with reaction time and button pressing, not to mention the clicking on minimap/unit control.

That isn't the point, that also is stupid. It isn't about whether or not we can make an AI that has the same detriments as humans as some sort of existential crisis solution.

That is an inherent property of being an AI playing a computer game.


okay im going to respond

no it isn't stupid as it adds one of the most important part of decision making in the game
Member
Posts: 25,903
Joined: Mar 11 2016
Gold: 17,609.00
Feb 12 2019 11:57am
Quote (Cpt_Ghost @ 12 Feb 2019 12:54)
okay im going to respond

no it isn't stupid as it adds one of the most important part of decision making in the game



computer can make decision faster than human.



Go Back To Starcraft I & II Topic List
Prev123Next
Add Reply New Topic New Poll