AI Olympics (multi-agent reinforcement learning)

These are 5 identical artificial intelligences. They aren’t very intelligent yet… But we’ll give them each a body and make them race until someone wins :) They need to learn to run 100m and they have 60 seconds to do it. The winner gets a cake :) Let's see what strategies they develop!

Attempt 1: Their movements start off random, but they're rewarded for moving forward and they're punished for falling over. So they're all in a lot of pain right now. .

. But they tweak their brains after each attempt, trying to maximize their rewards, and make it 100m! Like Purple!

You're kind of flopping around like a worm. . .

but it's a start, I suppose. Yellow is the first to learn to stand! I would recommend the rest of you do the same.

Purple, not only are you still flopping around, but now you’re going backwards too? While you’re embarassing yourself, Yellow is taking their first steps. Yellow, you’re still the only one able to stand has your walk improved yet?

You’ve definitely gotten better, you’re the first to make it to 10m! Green, you’re taking your first steps! You’re going the wrong way, but it’s a start.

And everyone has fallen over. Except for you, Yellow! Your walk is more of a scuttling right now, but you’ve passed your personal best and you’re still going strong.

Congratulations, you’re the first to reach 20m! It seems Red’s strategy so far is just to fall forward. .

. That’s not going to win you the cake, Red. Your punishments are based on muscle fatigue; so when you’ve all finished learning,fatigue; your movements should be looking human.

fatigue; uhhh, Green. . .

It’s hard to balance with one or two legs, so of course, Green and Yellow have an early lead. But both of you are quite slow, so we’ll see how long your lead lasts. .

. But for now, Green has a new best of 10m! Blue and Purple, you’re taking your first steps!

Well, I guess Purple is taking their first steps. . .

Your gentle hops seem to be doing very well! Not very fast yet, but it’s a promising start :) I guess Red’s decided to skip, meanwhile Purple is trying to cheat. .

. See Purple, cheaters get a face full of track. Purple, your hops are improving fast!

You’ve passed your personal best and are the first to make it to 40m! Even if it was just barely. .

. Red, you’re learning how to balance! And now you’re off the track.

. . Your skipping is getting more consistent, Red!

Nevermind. . .

I guess being tall makes it hard to balance. But Green, you’ve also been improving! Your shuffle is surprisingly consistent, having three legs makes it much easier to balance.

You’re flying past 20m! Meanwhile the others are still struggling. .

. All the way up to 50m, which means you’re in the lead now, Green! Why are you twisting so much, Blue?

It’s wobbly, but I guess it’s keeping you balanced. Meanwhile Purple is making huge strides! You’ve left everyone else behind!

You’ve passed the 60m mark. . .

and the 70m mark! That puts you in the lead, Purple! Are you going to have another go, Red?

Ah, still a bit over-enthusiastic I see. . .

No need to take it out on Purple! Look, now the others had to get involved. Red, your balance is almost perfect!

Though you’re still far from Albert’s skill level. You’ve definitely gotten a lot faster though! You hit 60m, a new personal best!

Not an inch more, of course. . .

Attempt 1000, a big milestone. That’s a lot of attempts, so let’s see where we’re at! Red still trying to escape the track, as usual Blue and Purple still can’t balance consistently, I guess it’s difficult with so few limbs.

Green and Yellow are still consistent. Consistent, but terribly slow. .

. You all need to pick up the pace. .

. This is a shockingly good start for you, Red, Though competition is close on your tail. Okay well maybe not Green.

. . Green, please, you’re making a fool of yourself.

Well Red, your tiptoeing has you in the lead now! You haven’t fallen off the track yet either. .

. There it is. Purple come on, it’s not nap time.

Red and Purple, what happened to your 70m runs? Are you trying to frontflip now Red? Now you’ve messed everyone up again.

. . Well except you Blue, is this your run?

Just steady up and you should be. . .

wait, where are you going, Blue? . .

. See you later I guess? Red, please, we’re running, not doing gymnastics.

Purple, your hops are looking really good! Actually, everyone seems to be doing pretty well You’re still going strong Purple, is this your run? Purple, you’re so close to the finish!

Just not quite fast enough yet. Another impressively strong start, Red. Purple too, we could have a close run!

It seems I spoke too soon. Well, Red, your leaping technique is quick! It hasn’t been the most consistent, but maybe this run will be different.

Well done, Red!