> A modern day version of Deep Blue vs chess champion Garry Kasparov.

Not really, the results will be uninteresting as OpenAI already crushed Dota 2 while handicapping their AI... 8 years ago

https://en.wikipedia.org/wiki/OpenAI_Five

It's also worth noting, DOTA 2 is far-and-away the more complicated game than League of Legends. A knowledgeable DOTA player has more tools to skill-gap their opponent than a professional League player does. The skill floor for most players is much higher on DOTA than League.

TAS has been out-microing humans since the Starcraft days, it's all but guaranteed T1 will lose by a sizable margin.

> It's also worth noting, DOTA 2 is far-and-away the more complicated game than League of Legends.

Although OpenAI Five didn't actually engage with DOTAs full complexity at any point. In its most advanced form the AI and human teams were still restricted to a pool of 18 fairly simple heroes from the full roster of 117 at the time, and several other standard mechanics were still disabled or simplified, which significantly narrowed the scope of what the AI had to deal with.

Good point, the roster is one of DOTA's difficulty sticking-points so that dulls the challenge quite a bit.

Nonetheless, while I cannot tell anyone to bet against Faker in good conscience, I still think T1 is screwed.

They also made the AI do vision processing iirc

Give them superhuman perception and it's a different ballpark. They win just by click rate / accuracy and coordination through a shared brain

Nah, OpenAI Five directly interfaced with the game through the bot API. It did have superhuman perception, they just artificially nerfed its reaction time to try and counterbalance it.

https://www.reddit.com/r/DotA2/comments/bf49yk/comment/elauw...

Maybe I'm remembering a different game ai accomplishment, starcraft perhaps

now that you mention that specific nerf, I remember that being a big one