AI learns to WALK 3D (Part 2)
Get an Exclusive NordVPN deal here: https://nordvpn.com/CodeBullet It’s risk-free with Nord’s 30-day money-back guarantee!
WATCH PART 1: https://youtu.be/qvpXpCvkqbc
Twitter: https://twitter.com/code_bullet
Patreon: https://www.patreon.com/CodeBullet
Discord: https://discord.gg/UZDMYx5
Art created by @Dachi.art https://www.instagram.com/dachi.art
Thanks again to NordVPN for sponsoring this video. Get an Exclusive NordVPN deal here: https://nordvpn.com/CodeBullet
It’s risk-free with Nord’s 30-day money-back guarantee!
Touch
Still waiting for the AI
You could try a negative point on reinforcement learning when the target is not in direct raycast of the “eyes” from the start then the learning would have adapted to forward walking instead of backward walking. 🤣
Like eyes on the prize that will force the AI to face the target and thus slowly stand up as well.
16:20 cursed angle (penis thumb)
thanks again for ****ing my ***damn ears 10/10
I think the proprioception idea is a good one, but I don’t think its enough alone. In us humans, that sense is combined with the inner ear (which conveys a sense or roll, pitch and yaw) and vision (which provides directional targeting). So along with a sense of proprioception, you need to simulate an inner ear and a “cone of vision.” Alas, you would have to train AIs to this task incrementally though. For example – you’d have to start training vision first (where it looks for a target in its cone of vision, and if it’s not there, it turns its head/body until it finds it). The inner ear should already be working for this part of the training (so a sense of perceived direction and range emerges). And once it’s learned to look at a target, then you can then try to train them to walk. As long as they learn to rotate themselves to look at the target first – they should walk forwards towards it.
Moral of the story? They should look before they leap. 😀
Aka make it learn to stand then look at something
PS. The reason I recommended the cone (instead of trying to calculate a volume of space and identifying objects within it for example) is because I believe the cone approach would simplify the task. If it’s a game object (even if it is invisible and non-clip) you can detect when it interacts with other objects. So the logic reduces to:
If vision cone interacts with target, then target has been seen.
But the cool part is, anything in the environment that interacts with that cone the AI can “see.” So it can be aware of and react to more than just the target. 🙂
@themonkeyman2547 I think it was being kind of efficient in its own way… I mean, it technically didn’t exert any effort to get moving, it just fell (backwards) in the right direction and then applied just enough force to keep it stumbling instead of falling down. LOL
@Thedarkbunnyrabbit But having less movements doesn’t necessarily always means consuming less energy. For example, jumping with one leg only probably require less movements than using both legs to walk, but it’s less energy efficient than just walking like a normal person.
@themonkeyman2547 I was thinking rewarding it for minimal movements made to get to the target would be good to minimize the random flailing. Penalize it for moving away, Penalize it more for falling. Reward it for moving toward the target, reward it more for having its chest face the target, reward it even more the lower the number of moves it has to take to reach the target.
Seeing the time lapses of the AI in their individual cubes feels disturbingly dystopian
@Agent paper turk er urbs
@Dolat Dey turk er yobs!
@Stupit the next logical step
@Stupit dear god AI is taking scammers jobs too hahaha
@Dolat AI learned to scam
Part three? We still need to see them walk like humans.
If he added the chest direction to target or limited the head range of motion I think it’s doable
@outsider344 pain is the first one I thought of. Walking like that would be uncomfortable. Although it’s still funny to see them try
Yes and also jump. i want to see them jump over obstacles
I was thinking part 3 would be obstacles, but seeing the video, yeah that’s not a bad idea
@Lord Yoofy The hell? Don’t you just turn into a gluon and quark soup then quantum tunnel to coordinate A to coordinate B?
Dude! Two videos in less than a week! Oh my GOD. That never happens. Looks like the AI that writes the script for the videos has been fine-tuned quite a bit…
I have a feeling you watch spiffing Brit
At 0 hours of training, the AI behaves remarkably like a newborn infant. The movements are random because it has yet to develop “awareness” of its limbs and how to use them. I just thought it was a really interesting comparison.
@Keyser Soju I bet it would eventually learn to use some glitch in the physics system to gain a huge amount of speed
I also trained my kids to walk by punishing them if anything but their feet touched the ground.
So from the whole beginning I was just an AI
@Keyser Soju I doubt it. Newborns move on to walking eventually because crawling is more tiring, and hurts the hands and knees after a while, and also probably because they see everybody else walking. Without the punishment for parts other than feet touching the ground to simulate the discomfort of the first two reasons, it would not have any reason to stop crawling
Imagine if he just left the crawling AI to learn for generations, I wonder if any of them would’ve evolved to walking like how a human learn to crawl first then learns to walk.
Give them a reward for moving in the direction of their front, a reward for keeping their shoulder height up (but penalized for shrugging XD), and a penalty that increases the longer their toes are on the ground without their heels also touching the ground, and make them have a vision cone for finding the target they have to move to 😀
Your comment made me realize this is pretty much exactly what pain does for humans.
Maybe penalize the AI slightly any time it overextends or over taxes its joints? Cause that shit would hurt if a human tried it.
@goldfndr Yes, that worked for me. I tried to walk like them and my neck started hurting after 5 seconds, so I turned around.
What about friction? Wouldn’t it be easier?
No need for a reward for moving in the direction of their front. Instead, penalize exponentially for head pivoting away from neutral.
perhaps the toe stance discouragement could begin if there is 1 to 3 seconds of toe touch without a heel touch (possibly also the reverse to discourage heel walking too) making the foot a little rounded in the back and with a flexible front section (toes) might also help to better match real feet. as for rewarding standing, reward head above shoulders, shoulders above torso, torso above legs, upper arm above lower arm, upper leg above lower leg, lower leg above foot (probably by failure hurting score, rather than success helping it much)
I wonder what would have happened if the “touching the floor” punishment were a reduction in score instead of immediate failure…
Rolling around at the speed of sound
Or if instead of a penalty, it was a reward? I want to see how fast they could worm 🪱.
i wanna see a part 3 where they walk like you know humans
3:42 – I’m saying though…This part was silly. I laughed so hard. But that sound track was pretty epic. Whoever produced that worked really hard. Probably was not anything easy about editing that part of the video either with the timing and all.
It went on for the perfect amount of time too! I couldn’t stop laughing