Seeing a lot of posts speculating that our results at The International show that today’s AI just can’t hope to match humans at strategy. Our interpretation is we have yet to reach the limits of current AI technology. Learning curves haven’t asymptoted just yet!

Aug 24, 2018 · 3:21 AM UTC

31
39
3
468
Note: would be quite a cool finding if our current approach does *not* scale to the level of the top human pros. But this week's matches were a snapshot of current progress, not what's possible.
10
5
1
86
Replying to @gdb
Please let AI ban-pick the heroes. These two drafts may be even from the view of HUMAN STRATEGIES, but it cannot be fair if human force their experience upon AI. It just missed the points of self leaning.
Replying to @gdb
I'm just a simple man, I see OpenAI didn't pick invoker, SF, Chen and Necrolic, my expectations went down.
1
Replying to @gdb
I really hope you guys are taking Data's strategy to winning into account when considering AI development. How well does the technology account for strategies not optimized for the best outcome, but to inhibit learning curve rigidity?
Replying to @gdb
Humans can enjoy winning AI for a little while, but it's temporary. AI intelligence is growing so fast, it's just a matter of time (basic math / graphing) to show that AI will indeed surpass the humans...and winning a game like this will become "child's play" for AI.
Replying to @gdb
It's impressive how far you've come, but I think open AI would stand a better chance if there was a draft phase.
Replying to @gdb
A bunch of code gave these guys at the top a fight. That in its own right is impressive.
Replying to @gdb
Have you published the learning curves of your last self training sessions? Eg frequency of winning against past snapshots with snapshots equally spaced in terms of computation time.
3
Replying to @gdb
Wasn’t that the point of choosing Dota2? To take on a concrete milestone you weren’t sure AI could handle now but that seemed reachable given current tech, now that AI can kick butt at the previous “can’t just hope to match humans” milestone of Go?
Replying to @gdb
If you succeeded every time, all it’d tell you was that you were playing it too safe