AI has mastered Gran Turismo
"An artificial intelligence program has beaten the world's best players in the popular PlayStation racing game Gran Turismo Sport, and in doing so may have contributed towards designing better autonomous vehicles in the real world, according to one expert.
The latest development comes after an interesting couple of decades for A.I. playing games.
It began with chess, when world champion Garry Kasparov lost to IBM's Deep Blue in a match in 1997. Then with Go, when A.I. beat Korean champion Lee Sedol in 2016. And by 2019, an A.I. program ranked higher than 99.8% of world players in the wildly popular real-time strategy game StarCraft 2.
Now, an A.I. program has dethroned the best human players in the professional esports world of Gran Turismo Sport."
Took a couple of decades to figure out closed circuits. Now let's see how it fares in a proper rally.
In reply to Kreb (Forum Supporter) :
On the one hand that's an excellent point.
On the other, depending on how it's allowed to make its pace notes, it might be able to do that near-perfectly.
None of which is as complicated as driving in public, though, is it? Which I still feel like we're involuntarily beta testing.
The most interesting part of this article to me was this:
"I learned a lot from the A.I. agent," Miyazono said. "In order to drive faster, the A.I. drives in a way that we would have never come up with, which actually made sense when I saw its maneuvers."
I can't help but wonder if F1 teams can leverage this to exploit lap times at tracks. Could the AI find a line or a passing advantage that the drivers don't see but they could then copy it?
In reply to NY Nick :
That jumped out to me, too.
NY Nick said:
The most interesting part of this article to me was this:
"I learned a lot from the A.I. agent," Miyazono said. "In order to drive faster, the A.I. drives in a way that we would have never come up with, which actually made sense when I saw its maneuvers."
I can't help but wonder if F1 teams can leverage this to exploit lap times at tracks. Could the AI find a line or a passing advantage that the drivers don't see but they could then copy it?
I noticed one of the fast guys in my karting league was trying all kinds of crazy lines last night.
I'd struggle to know if the result was actually faster or not (or even to consistently pull them off) but I have a feeling that some of his speed comes from his creativity in that department.
So, an imperfect racing simulation that relies on 1's and 0's was mastered by a computer that sees those 1's and 0's and where they can be exploited? Color me shocked.
I remember an interview with Hans Stuck talking about a factory prototype he was running at Spa (it's been a while since I've heard this story during a break in some race, so some details might be fuzzy). The lead engineer was adamant that Stuck could not run a lap as fast as he did in qualifying, because it was a tenth or two faster that what their computers said was the fastest lap possible. The engineer called him up at three a.m. on race day to tell Stuck that he was flat out in fourth gear through a section that the computer was still in third for to keep the car more stable.
