Truly, AI is just here to be our friends.

​@08weavn I bet if they improve this tech they will sell it to the military

​@@Durrtyboy The AI will now pick strong or smart men and women in the military, instead of picking weak ones. No more adding liability.

@@cye2310 Nah, smart humans are a liability to the AI. It would be smarter for it to surround itself with mindless grunts that do nothing but obey orders, The last thing it wants are humans that are smart enough to avoid going on suicide missions for the greater good.

The military already has A.I. doing strategy btw, they have some pretty cool tech that is public knowledge. But the final decisions always go to humans except for close defense systems which obviously don't have time to be reviewed and there's no repercussions if they misfire. It's important to note the distinction between general A.I. and the current A.I. models that we have. What kirbyjoe said is general A.I. which does not apply to this or chat gpt, or your common prompt based stuff. These A.I. are trained for a specific task through machine learning, that's all they will ever do. They inform us on how to program more efficiently and help us learn about them for other purposes, but all they do is play SC2. The same way generative A.I. just takes something humans did and makes an answer to it. Meaning if humans stop doing stuff, they just get stuck in time. General A.I., meaning A.I. that self teaches and evolves similar to us humans is way way off despite what most people think@@Durrtyboy

Yeah, that makes a lot of sense. Google's Alpha bots had to play the equivalent of thousands of hours, if not decades worth of games to reach professional-level decision making and plays iirc. I can't imagine the botmakers here have access to the server power for that sort of intensity haha. Diamond is still really impressive though! Compared to any of the AI in Starcraft, even with cheats, these AI are clearly better without access to any cheats. Am I also correct that the messages the bots type out are pre-programmed messages made by the bot programmers? I imagine they're useful for debugging, essentially.

My favourite analogy is, imagine every unit being micro'ed by a pro player, but all the decisions are being made by a bronze-league player. That is how bots average out to diamond'ish league level.

@@Ketrocindeed, perhaps there should be an AI league with ai capped at around 800-1000 APM? Would this force developers to build intelligence?

@@basicuser54 It's not like the bot developers are intentionally making the AIs stupid, it's just that it's really really hard to code an AI that handles the general case intelligently. You can make an AI that plays well in a few specific scenarios, but as soon as something unexpected happens the AI has no idea how to respond, and it's incredibly difficult to fix that.

@@basicuser54 Decisions in assessing a situation is what bots are poor at. Lowering apm would mostly just be a process of removing redundant commands and grouping units who need a similar command... So it'd mostly just be making more overhead for authors in terms of micro. If our bots ever become too strong for humans then maybe limiting apm for fairness would be a good idea.

Yo, it made me wonder if the adjutant was ever allowed to order the terran troops whenever the commander was not around

​​@@hansmatos2504I mean in the Campaign Raynor asked the Adjutant if his troops were ready. So I guess they manage troops, it wouldn't be far fetch if they can command in battle.

i wanna see Lowko cast a Z vs Z Ai match ... XD

Unless you have 20k APM, there's very little to be learned.

exactly. if they strategy is based on constantly clicking on workers and rerouting them... then there isn't much to learn... it might force it to do more inspiring strategies if we cap the apm.

Obviously he did not mean learn how to play. He ment learn how to program, understand the mechanisms these algorithms use for decision making....

I wonder what the thought process was behind sending marines after marines into their doom

AlphaStar was fun to watch, especially when beating pros, not this high apm mess ;)

I still remember Ketrok for the polanetary forotdds

He run one of the first industrial scale salt mines on the ladder.

@@paulsd9255 Happy LagTV is back? They got too hungry on just the podcast, I think. They've had some great casts lately though.

Also, Ketroc Bot randomly choose a strategy at the start of the game

This is bang on

With a game that's won or lost due to attrition, you would need to make better trades, mine more resources, or both. If you mule your opponents minerals, even if the mules don't manage to return those minerals to you, you still take them away from your opponent, so they might run out faster

Polanetary forotdds

Ketroc rocks

yup, its called speed mining! its a standard in bot meta!

I'm surprised I didn't see them using the bunker trick with the refineries - the AIs I've seen in the past usually put 2 SCVs on each refinery and a bunker (well, when the 2 geysers are both next to each other anyway) to teleport the SCV across the bunker and get full gas mining with only 4 workers instead of 6 (effectively using a bunker as 2 SCVs that don't take up the CC's build time or supply).

@@asdfqwerty14587 that's something you'll usually see with MicroMachine and only when the geysers are beside each other. It's not cost effecient otherwise so it's map dependant

@@ProBotsAI Good to know. I also came to the comments wondering why neither bot was bunker mining.

Damn

they do act retarded and there is little to no discipline n their playstyle.

i thought of this strategy before , it seems pretty damn good in some rare macro games where the map getting mined out , its good

The idea is, if the bot is going mass ravens and already has a mineral bank, it makes more sense to steal the minerals from the opponent. FYI, when Ketroc is confident it'll win, it creates animated pictures on the minimap with mules instead :).

@@Ketroc That is actually awesome! I'm happy fun stuff was included after all XD

@@Ketroc Animated? Wow that sounds awesome!

Zerg bots make burrow seem so OP. Imagine each individual roach in an army burrowing and retreating from a fight when they get low in health. Perfect creep spread and injects is insane too. I think toss/terran have more microable units in general though. Units with speed and range do well with bots for their kiting ability.

​@@Ketrocbut than again....there is no splash against perfect speedling micro. There are clips on ytb of AI splitting against multiple siedgetanks

@@Ketroc Isn't that more or less just bootleg Blink micro? Keeps driving me nuts that Zerg players don't use Burrow just to keep Roaches in good health. Could use the spared larva for drones.

@@stixinst5791 that must have been done in map editor. Bots can't split vs tank shots as they don't know what the tanks are targeting. They can space out everything though, like how mutas magic-box thors. I've seen marines space out to defeat tanks in bot games. Speedlings are nice as they can negate the opponent's micro. You can't really kite vs lings as they are so fast. So long as the zerg bot is smart enough to know when to swarm with lings and when not to engage, they can be very powerful.

@@Komatik_ it's like blink + cloak + your army won't body-block you + you heal faster than medivac healing rate. Much more actions than blink-micro though which is why humans don't use it.

you'll notice it definitely in the builds

That’s called single player mode, he’s playing single player mode versus the computer

Harstem already did it and the bot got destroyed, he thought it was about diamond level

@@AD-lr6pc In the other Lowko cast, the bot won

can you link that @@nemou4985

I would love to see some "Cheesiest man alive" against AIs

@@wernerviehhauser94 I don't think the AI is meant to be cheese-able, and that's why their natural is a Planetary Fortress. They always play very safe, maybe overly safe. I think it'd be better to see some "Greediest Man Alive" against AIs.

@@jingxuanwang8674 actually, now I want to watch both!

@@wernerviehhauser94 Well, it'd be a super easy challenge and a super hard challenge. Both sounds good to be honest.

The dancing is the bot staying out of shooting range of danger and trying to attack. Each unit of a bot can detect if the unit is in range of an enemy unit and tries to stay away but also is being issued commands to attack so they are bouncing back and forth from attack and safety.

@@ProBotsAI I can see that it’s moving away before the attack animation tick

have you seen how human Ketroc plays the game? it's exactly the same, but just on 90 apm :)

You've perfectly described bots. Micro/macro beyond pro level. Decision-making below bronze level.

Giant Grant Games played against AI from each race on his archive channel

Zerg with perfect creep spreading would be really annoying.

@@rafaeljosebragacoelho6217 I first thought Marines are a lot more micro-able than Zerglings but Zerg definitely have the best Spellcasters. Ghosts have EMP and can dodge Banelings. Zerg have legit infinite vision. I don't get how Stalkers are good though. EMP + Liberators (or Stim Marines) kills them all.

@@jingxuanwang8674 Stalkers with an ultra precise use of the blink,get you able to just squeeze their maximum, i dunno for real about the game meta ,but i see ghosts as a Double-edged sword while being extremely dangerous to deal against it, but costly and fragile as the game progresses. Talking about those bots, Ketroc won because it just sticked to a strategy and waited the other side spoil all of its mines, in the end it was a economical victory.

It's a fairly balanced meta for bots. Protoss can be good too. Tempests always kiting from max range. Phoenix can literally never take damage as they are faster than all anti-air units. Even something like feedback can be powerful in the hands of a bot. Zerg may have the least micro potential, but creep spread, injects, and burrow-micro are so strong. Current #1 bot on aiarena ladder is zerg.

2.7 mbs