> Many of the pros, even ones who have played against (read: got crushed by) OpenAI Five, still think it's very far away for the bot to out draft a pro team with the full ranked game rules and hero pool.
Neither AI experts or DotA experts have the necessary background to predict how close anything is to any level of future performance. DotA players are probably the worst to ask, because the AI is already stronger than they are at a subset of the game and they have no insight into how quickly that subset could be generalised into other aspects.
OpenAI has a list of something like 7 restrictions. Nobody has any real idea of how quickly these restrictions can be lifted once the OpenAI team has an AI that has mastered the game with those restrictions.
Eg, 5 invulnerable couriers is obviously a huge restriction - but once OpenAI knows it gains a benefit by ordering items, how difficult is it to lift that restriction? Nobody knows. Might be easy, might be hard.
> OpenAI has a list of something like 7 restrictions.
Some of those are very major, though, and I think the word "restriction" is a bit misleading. Having 5 invulnerable couriers is not really a "restriction" in that it limits or simplifies parts of the game -- it's just a fundamentally different mechanic that changes the way the game can be played.
> DotA players are probably the worst to ask, because the AI is already stronger than they are at a subset of the game and they have no insight into how quickly that subset could be generalised into other aspects.
I think that's a little unfair. Most folks have been pointing out that OpenAI's current momentum-based "deathball" strategy seems to fall apart without infinite regen and a limited hero pool, both facilitated by the current set of restrictions.
I'd agree that nobody really knows how well OpenAI will adapt to the full game, but I disagree that the criticisms I've seen are meritless. OpenAI's current level of play is definitely impressive, but I think there's still room for skepticism given the current restrictions. I (and I think a lot of others) would be pretty disappointed if the TI showmatch happened with the turbo mode couriers still enabled.
> I think that's a little unfair. Most folks have been pointing out that OpenAI's current momentum-based "deathball" strategy seems to fall apart without infinite regen and a limited hero pool, both facilitated by the current set of restrictions.
People said similar things about Go AIs and ko fights. And in the end it turned out that neural networks handled kos fine but ladders were a challenge.
On the deathball strategy in particular, consider that we expect a superhuman DotA AI to change the DotA metagame, so playing off-meta doesn't tell us anything. AlphaGo would invade 3-3 point a lot more enthusiastically than a human player. This was considered a classic beginner mistake for many years; now the theory has been readjusted to cope with the fact that AlphaGo stuck with it and just considered it a good move.
We can safely say that the courier change has made a deathball strategy more powerful and it seems quite likely it is not an optimum strategy. But we can't be sure until OpenAI tests it, and we absolutely can't be sure that OpenAI won't just learn a new style when the conditions change.
The criticisms have merit, but nobody has enough data predict anything about the future. Particularly a professional DotA player.
>I (and I think a lot of others) would be pretty disappointed if the TI showmatch happened with the turbo mode couriers still enabled.
Totally agree. This one change alone is _so central_ to both the bots laning strategies and meta-game team strategies. They can't just leave heroes in lanes forever no matter what, and have all 5 heroes literally never go back to the well if there aren't 5 couriers. Not to mention their initial item builds, stats-only-4-man-the-lane-for-first-blood bullshit wouldn't work at all without constant ferrying of regen on the couriers.
Neither AI experts or DotA experts have the necessary background to predict how close anything is to any level of future performance. DotA players are probably the worst to ask, because the AI is already stronger than they are at a subset of the game and they have no insight into how quickly that subset could be generalised into other aspects.
OpenAI has a list of something like 7 restrictions. Nobody has any real idea of how quickly these restrictions can be lifted once the OpenAI team has an AI that has mastered the game with those restrictions.
Eg, 5 invulnerable couriers is obviously a huge restriction - but once OpenAI knows it gains a benefit by ordering items, how difficult is it to lift that restriction? Nobody knows. Might be easy, might be hard.