Moving this video to RFlagg's personal queue. It failed to receive enough votes to get sifted up to the front page within 2 days.

*promote

Promoting this video and sending it back into the queue for one more try; last queued - promote requested by eric3579.

1v1 mid to 5v5 is like checkers to chess, so I'd be interested to see if it can also master all of that.

It's curious how aggressive it plays. When you watch AlphaGo I think one of the defining features of it was that didn't look for huge leads because the margin of victory is meaningless. OpenAI seems totally different, it practices a sort of brinkmanship that is almost human. The games I've seen it lose were because of moves that actually did seem very human - it went all in up high ground and got unlucky for instance. I'd have somewhat expected it to evaluate doing that as an unnecessary risk when it can just extend its advantage in other ways... but then maybe it is overestimating the ability of its human opponents to keep up with last hits and raze placement.

*future

Adding video to channels (Future) - requested by notarobot.

It's not an apples to apples competition. The bot is cheating because it has perfect knowledge; it can look bounding spheres and exact positions, maybe even knows the enemy cooldowns, etc. Basically it's using an aimbot that can always react faster than a human.

What they need to do is require the bot use vision, either only give it access to the frame buffer, or better yet require it to use a camera pointed at the screen. (some day I'd require it to use a mouse and keyboard)

It would be interesting if you slowed the game down 10x, to see if the humans would regain the advantage.

Dunno if I'd call that cheating. I noticed in the showmatch it seemed to adjust its creep block without any vision of Dendi's, but I think they said since that the bot doesn't know what it doesn't have vision of. I mean, I agree it has a massive advantage due to a sort of almost latency free interface with the game while us humans have to do all sorts of shit before we can even begin to make a decision, but then nothing about this competition is equal. You might just as well say it is cheating because it was able to play more Dota 2 in two weeks than any human could in their lifetime...

Does anyone remember the early days of bots in FPS games? I'm not impressed by the bot winning because of those early FPS bots.

Walk round a corner with an old fashioned Quake 1 bot? BAM you're dead as soon as a pixel of your head appears (headshots existed in QTF); you probably won't even see the bot because of your refresh rate.

The learning is incredibly impressive, but as someone said earlier the impressive thing would be filling in as a player in a 5v5, or maybe a team of them. I'd always expect a bot to win a duel of reactions unless they are programmed to miss. The bot makes its decisions and performs them instantly, so if you make better decisions than it, they still must be better enough that you can beat that huge speed advantage.