OpenAi Creates Bot That Wrecks Pro DOTA2 Players

1 Like

I can’t wait to see the 5v5 next TI with the bots. It’s actually so damn ambitious. And if they succeed… guys, listen ok, listen closely. We are all absolutely screwed if the bots win that 5v5. Probably. Though it will prob require even more direction/input from the developers for the bots to stand a chance.

Machines have better ways to handle things. We handle things with our senses and our hands. Machines have senses which produce numbered data to calculate with precision. And they don’t need hands, they can have things like guns optimally attached to like a robot. So when you compare machines handling weapons, and humans handling weapons, it will eventually always be no contest, in the machines favor. As you can see in DotA where the weapons are a mouse and keyboard vs the bots direct inputs. The only thing that was missing from the bots was decision making capability, and now they’re closing that gap. Anyways… dont actually worry about it?

Always exciting to see these developments in AI.

DotA sounds like a good proving ground for development of these techniques given the complexities that have to be taken into account, the strategies that need to be planned and the degrees of freedom that need to be controlled.

If they are reaching the point where they can start to understand the underlying complexities of a game like this and reliably work out and implement optimal solutions, such as beating 8 pro players in a row.
I can’t imagine what these techniques would achieve when they are let loose on real world logistical problems.

  • How does a government plan its’ budget to minimise hardship of its’ citizens
  • How does large scale farming adjust its’ crop selections and placements to achieve the best outcomes from predicted weather in the coming season and dietary trends
  • How can traffic light timing or Lane control be altered in real time to mimimise wasted time in traffic
  • How can AI be used to make better AI

Insha Allah exciting times ahead.

They beat all those pros, but only 1v1 mid shadowfiend. Also to the question of how AI can be used to make better AI, that’s actually exactly how the bot learned, by versing itself, simluated and processed over a 2 week period I think? Might be wrong on how long they processed for. What I remember clearly from a different video about it is the bot had “lifetimes” of experience by the time it was put to the test against the pros.

I think if this isn’t regulated from the get go, it’s not gonna be good. Problem is, I don’t think regulation can work for something like this. It’s like trying to regulate or control nukes or chemical weapons, which are infinitely easier to control than AI advancment, and that didn’t work so…