Page 1 of 4 1 2 3 ... LastLast
Results 1 to 10 of 38

Thread: open-ai's 1v1 mid bot code open source?

  1. #1

    open-ai's 1v1 mid bot code open source?

    Is it going to be possible to see the source code of the machine learning stuff.

    Im pretty sure it'd be great to learn from and be really helpful for most people here

    I had imagined the ML-stuff for laning/1v1 mid would just focus on walking to lane...then do the laning algorithms it learns on. but it teaching itself to creep block like that is pretty impressive and cool

    maybe I should be asking this question directly to the open-ai people
    https://github.com/ThePianoDentist/t...dentistdotabot Lina bot which pulls small camp when 'laning' (Aim to work on pulling and stacking bots initially)
    https://github.com/ThePianoDentist/dotabots-ml-tools Parsing data from bot games

  2. #2
    Basic Member
    Join Date
    Dec 2016
    Posts
    76
    maybe, I really want to know how they did it. since i failed to do the same thing
    https://github.com/lenLRX/Dota2_DPPO_bots ----My ML bot work in progress

  3. #3
    poor dendi...^_^

  4. #4
    10+ people with a lot of money wrote a 1v1 SF mid but after a couple of months that wins if enemy obeys their rules AND only picks SF, honestly, nothing to see here! I'm sorry, but I also don't buy the "it started with nothing but random moves and learned everything" BS (it is a good bait for getting fundings from clueless people tho). And don't bother asking for its code, I don't think the code will ever be public for obvious reasons.

    The thing that bothers me the most is that Valve got fooled with this BS, but didn't bother to run an actual 5v5 bot vs bot tournament for the people who are actually making bots.

  5. #5
    Basic Member
    Join Date
    Oct 2016
    Posts
    247
    yea it made me sad that they didnt run any workshop bot code.

    but if they do develop the 5v5 like the guy said..... i hope it never becomes playable in the workshop or else all our efforts are for naught.... the API is prety much for naught..... those bots would make all other obsolete - especially if it learned to play "fair" and difficulty settings worked properly.

  6. #6
    Basic Member
    Join Date
    Dec 2016
    Posts
    46
    Quote Originally Posted by Platinum_dota2 View Post
    10+ people with a lot of money wrote a 1v1 SF mid but after a couple of months that wins if enemy obeys their rules AND only picks SF, honestly, nothing to see here! I'm sorry, but I also don't buy the "it started with nothing but random moves and learned everything" BS (it is a good bait for getting fundings from clueless people tho). And don't bother asking for its code, I don't think the code will ever be public for obvious reasons.

    The thing that bothers me the most is that Valve got fooled with this BS, but didn't bother to run an actual 5v5 bot vs bot tournament for the people who are actually making bots.
    I do not know if this AI is written by lua. Even if our script is not very strong, but we should get more publicity. And If valve chose to ignore our AI script developers, I would feel some disappointment.

  7. #7
    Quote Originally Posted by Yavimaya View Post
    yea it made me sad that they didnt run any workshop bot code.

    but if they do develop the 5v5 like the guy said..... i hope it never becomes playable in the workshop or else all our efforts are for naught.... the API is prety much for naught..... those bots would make all other obsolete - especially if it learned to play "fair" and difficulty settings worked properly.
    Don't worry, if at any point they wrote a team of bots only with the API that is available to us (and without any made-up BS constraints), I'm up for the challenge to beat their bots' asses with my bots (and I'm not talking about PubSimulator which I wrote in the last spring break!). However, I don't think this will ever happen and they will probably switch to another easy but eye catching project that makes them look good (1v1 LoL maybe?).

    But yeah, Valve managed to exteremely piss me off with this segment, this could easily be the final match of a 5v5 bot vs bot tournament instead of a coverup for the Elon Musk's $1b failure and making their players look like fools in the outsiders' eyes. Extremely short sighted!

  8. #8
    the problem is 5v5 goes on for too long to be a nice watchable segment at TI.

    whilst points raised about downside of it are legit, you cannot argue it wasn't a decent to watch segment imo.
    https://github.com/ThePianoDentist/t...dentistdotabot Lina bot which pulls small camp when 'laning' (Aim to work on pulling and stacking bots initially)
    https://github.com/ThePianoDentist/dotabots-ml-tools Parsing data from bot games

  9. #9
    Quote Originally Posted by adamqqq View Post
    I do not know if this AI is written by lua.
    I also really doubt that. Checking unit animations was only recently added and we still can't aggro creeps properly with the current API and their bot was using both of these.

    Quote Originally Posted by adamqqq View Post
    Even if our script is not very strong, but we should get more publicity. And If valve chose to ignore our AI script developers, I would feel some disappointment.
    As I said, the OpenAI thing was just a show, nothing more. They wrote a bot for a single hero with basically 1 spell that only goes mid and can only play against his hero! This is basically like writing a bot that plays a version of chess where players only have a single soldier! In what world this is better than the current workshop bots?

    I also hope Valve announce something like a 5v5 bot vs bot tournament soon, otherwise, they have to pray that these folks write actual bots for them (which I highly doubt)!

  10. #10
    if it doesnt get open sourced that would be disappointing though. Im pretty sure if you're gonna call you're company Open<something>....it should prob be open source

    so do you think it's using image processing rather than in-game variables?
    I thought it would be through in-game lua scripting because I dont remember anyone requesting opening up CreateHTTPRequest to non localhost urls. But that is something that's necessary for them to run their bot, I could be wrong though.

    also Im ok with what they did, I think its pretty cool.

    I think making statements like "OpenAI first ever to defeat world's best players in competitive eSports. Vastly more complex than traditional board games like chess & Go." is a bit dramatic though. Way too clickbaity for my liking.

    Considering it took 2 weeks to train that well. you've got 25x more time for 5v5. 25x I consider not a lot when a) the games last 5x as long. b) you're going to have orders of magnitude more input features. We'll see how well they can do.
    Last edited by TheP1anoDentist; 08-11-2017 at 09:07 PM.
    https://github.com/ThePianoDentist/t...dentistdotabot Lina bot which pulls small camp when 'laning' (Aim to work on pulling and stacking bots initially)
    https://github.com/ThePianoDentist/dotabots-ml-tools Parsing data from bot games

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •