August 17th, 2019, 02:16
(This post was last modified: August 17th, 2019, 02:17 by Flagris.)
Posts: 8
Threads: 0
Joined: Aug 2019
(August 16th, 2019, 05:31)Blake00 Wrote: Many of the reviews (even the bad ones) all seem to talk about how hardcore the AI is so that's promising.
Nah. It is not promising because Pandora AI in its present state is completely fan-written. Proxy Studios (Pandora devs) are ruining now the WH40K universe with their Gladius game (based on Pandora engine). These guys can't make a decent AI. As well as a good and original game at all (Pandora is a cheap and shallow rip-off of Sid Meier's Alpha Centauri).
And Slitherine don't develop games no more, now they are publishers only. And most of their published games have big problems with quality, alas. Broken promises, lack of support, developers abandoning games and so on.
As for 'cloud' subject, it would be another NAY-factor for me personally. Everything external just means my giving up of control over my games, be it 'clients' like Steam one, 'cloud AI' or Diablo 3 'online only' gameplay. It's risky and with very high probability (i.e. - almost always, if not always) leads to various kinds of failures, 'end of support', 'shutting the servers down', 'due to diminishing of players base... blah-blah' and all other similar quibbles that just mean the same: "Spice must flow".
Posts: 542
Threads: 4
Joined: Jul 2017
(August 16th, 2019, 18:55)haphazard1 Wrote: I think a modern 4x game is rather more complex than chess. ....
Edit: Forgot a huge one.  Chess ia a game where all players have perfect knowledge. The comparison was with StarCraft, why are you comparing with chess that was solved 15 years ago? StarCraft also has fog of war, and a harder one to boot - you don't see structures in it while you see mom explored cities in their current state.
(August 16th, 2019, 15:59)Seravy Wrote: I guess we'll know who is right in a few years. In the end, we are both guessing on the complexity. If it is possible to use that type of AI for 4X already, someone will sooner or later do it and then it'll be obvious. No, I am guessing on the search space dimension. (complexity is not the point, see chess vs go) You're grasping at straws and didn't even know any of this till 2 days ago. The point is that it's probably already possible but even if it weren't, it'd be in a matter of months.
(August 17th, 2019, 02:16)Flagris Wrote: (August 16th, 2019, 05:31)Blake00 Wrote: Many of the reviews (even the bad ones) all seem to talk about how hardcore the AI is so that's promising.
Nah. It is not promising because Pandora AI in its present state is completely fan-written. Proxy Studios (Pandora devs) are ruining now the WH40K universe with their Gladius game (based on Pandora engine). These guys can't make a decent AI. As well as a good and original game at all (Pandora is a cheap and shallow rip-off of Sid Meier's Alpha Centauri).
And Slitherine don't develop games no more, now they are publishers only. And most of their published games have big problems with quality, alas. Broken promises, lack of support, developers abandoning games and so on.
As for 'cloud' subject, it would be another NAY-factor for me personally. Everything external just means my giving up of control over my games, be it 'clients' like Steam one, 'cloud AI' or Diablo 3 'online only' gameplay. It's risky and with very high probability (i.e. - almost always, if not always) leads to various kinds of failures, 'end of support', 'shutting the servers down', 'due to diminishing of players base... blah-blah' and all other similar quibbles that just mean the same: "Spice must flow". Even if cloud AI solved the issue related to AI that you agree with me about in the first part of your post?
I mean, you'll agree that there are two conflicting priorities here.
... So what if the game service was delivered by the same company offering the AI cloud?
https://www.polygon.com/platform/amp/201...nouncement
The future of gaming is quite clear, i hope mom fans don't miss out on it.
Posts: 8
Threads: 0
Joined: Aug 2019
(August 17th, 2019, 03:50)Arnuz Wrote: The future of gaming is quite clear, i hope mom fans don't miss out on it.
O yes. It's already here, this future. There are no more good old 'computer games' as we knew it. And I plan to miss it out completely.
Posts: 114
Threads: 1
Joined: Nov 2014
IMO Slitherine should hire Seravy and a good player of 'Magic the Gathering' to build a new and perfect 'Master of Magic 2' ;-)
Posts: 114
Threads: 1
Joined: Nov 2014
@Arnuz: To build an AI is more then a cloud with agents. The AI for Go and Chess needs years of development with great teams - and this are games have easy rules and a few informations (haphazard1 lists it).
If AI would be easy (or only doable), why we don't have one to solve the problems of our time? ;-)
Posts: 10,492
Threads: 395
Joined: Aug 2015
Quote:No, I am guessing on the search space dimension. (complexity is not the point, see chess vs go) You're grasping at straws and didn't even know any of this till 2 days ago. The point is that it's probably already possible but even if it weren't, it'd be in a matter of months.
Whatever, my point is, if no one comes out with a good 4X AI in the next, let's say, 3 years, you'll have to admit you were wrong. You say months, so I'm being generous.
Posts: 542
Threads: 4
Joined: Jul 2017
(August 17th, 2019, 08:14)vicwaberub Wrote: @Arnuz: To build an AI is more then a cloud with agents. The AI for Go and Chess needs years of development with great teams - and this are games have easy rules and a few informations (haphazard1 lists it).
Except, when experts do that - who understand the problem as they've only been working on AI - it takes three days:
https://www.nature.com/nature/journal/v5...24270.html
Or a simpler article about it: https://edgy.app/new-alphago-program-is-...ld-version
That's the evolutionary jump in AI I'm talking about.
Clearly starting to do 4x games will offer its own challenges. The question is: would these challenges be easier to overcome for the guys who beat StarCraft pro players or for fucking Pandora devs?
... I know where my money is.
Posts: 114
Threads: 1
Joined: Nov 2014
(August 17th, 2019, 08:36)Arnuz Wrote: (August 17th, 2019, 08:14)vicwaberub Wrote: @Arnuz: To build an AI is more then a cloud with agents. The AI for Go and Chess needs years of development with great teams - and this are games have easy rules and a few informations (haphazard1 lists it).
Except, when experts do that - who understand the problem as they've only been working on AI - it takes three days Three days for a game with 19x19 fields and only 4 rules. It is not a jump I think. As I know it based on theorys of the 1950s combined with a big hardware base and a database of 30 millions human moves.
Does anybody know about Deity Empires? It's a new successor of MoM under heavy development.
August 18th, 2019, 07:20
(This post was last modified: August 18th, 2019, 07:27 by Flagris.)
Posts: 8
Threads: 0
Joined: Aug 2019
(August 17th, 2019, 08:36)Arnuz Wrote: That's the evolutionary jump in AI I'm talking about.
https://www.wired.com/story/deepminds-lo...elligence/
Summing up what is written in this article, “How do children acquire language and come to understand the world, using less power and data than current AI systems do?” If we spent more time, money, and energy on the latter question than the former, we might get to artificial general intelligence a lot sooner.
They just have to realize that human-like intellect is inseparable from the whole human constitution (physical factors, genetic/biological factors, social factors, cosmic factors, spiritual factors). As long as they treat mind as an independent, simple mathematical function - they go nowhere.
Though I doubt they really want to. As soon as I see/hear ' We need MORE money for this/that', I turn around and walk away laughing. It's just like believing in Elon Musk's "genius" or something similar.
Posts: 10,492
Threads: 395
Joined: Aug 2015
Nice article. I guess it pretty much proves my point.
Quote:DeepMind’s StarCraft outcomes were similarly limited, with better-than-human results when played on a single map with a single “race” of character, but poorer results on different maps and with different characters. To switch characters, you need to retrain the system from scratch.
So basically it didn't learn to play the game. It memorized the optimal strategy for an already known and explored map and enemy. That's not yet usable for a game AI, Starcraft or 4X doesn't even matter.
Quote:By one estimate, the training time for AlphaGo cost $35 million; the same estimate likened the amount of energy used to the energy consumed by 12,760 human brains running continuously for three days without sleep.
In other words no commercial game company can afford it even if it was capable of playing the game. Even the best selling 4X games don't make $35 million, do they?
|
