Profile picture
Smerity @Smerity
, 11 tweets, 6 min read Read on Twitter
Congratulations to the @OpenAI Five team! Whilst the bots may have lost we've all gained from watching your journey :) My live stream commentary is linked and below I'll add a bit of post game commentary.
twitch.tv/videos/3011031…
First, we should all note how impressive it has been to watch the restrictions removed in the last two or so months. The largest was from a fixed set of heroes to a small hero pool but there are many on top of that. IMHO the game is now functionally equivalent to normal gameplay.
The addition of a vulnerable courier was an enormous change that I haven't seen much commentary on. This courier can be seen as the sixth player of @OpenAI Five, responsible for ferrying items to you through no man's land. Poor little fellow needed a longer bootcamp! :'(
The bots still had trouble using wards which give vision. Wards were overlapping and frequently placed in range of their own tower (useless as the tower already provides vision both of visible and invisible units). Another result of the complications due to less restrictions tho!
I also think I saw some fairly major bugs that likely resulted in the loss. Axe (big scary guy w/ an axe 😅) has an ultimate called Culling Blade. If the enemy is below a certain HP they die instantly, you move faster, and can use Culling Blade again. He flubbed more than once 😭
Necro had similar issue. His Reaper's Scythe deals damage according to how much HP was missing yet he used it very early in many fights. Even more important, if the character dies from Scythe, their respawn time is +15/30/45 seconds. Axe +Necro had more than one instance of this.
Having said that,
- @OpenAI started this bot training run fairly late
- @OpenAI removed _many_ restrictions at the same time (+increased likelihood of bugs / bad behaviour)
- @OpenAI will have worked out most of the bugs and .:
- @OpenAI should be very confident for next year :)
As opposed to AlphaGo, @OpenAI Five still has only used self play. As someone on my stream noted, many of the behaviours these bots need to learn are not easy from self play. There's a treasure trove of millions of human DotA games they could also learn from.
Finally, these models can run on your laptop. There's a gaming community of millions of DotA players who could benefit from it. There's a million interesting research questions that this work could result in. Robotic arms use the same backend. I can't wait to see what's next! ^_^
On a personal note, DotA1 was why I ended up on Twitter with this being one of my first ever tweets (and one of ten in all of 2008) ...
(tldr we coordinated university students to play DotA by Twitter SMS 😅)
and even pondered in 2016 with @roybahat and @jackclarkSF (who I think was still at @business at the time?) about DotA / LoL being an amazing proving ground for AI. @OpenAI proved us right - and there are still many mountains to climb ^_^
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Smerity
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($3.00/month or $30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!