AI Learns to Rocket Jump

AI Learns to Rocket Jump

base_thomas

4 месяца назад

577,370 Просмотров

Ссылки и html тэги не поддерживаются


Комментарии:

@base_thomas
@base_thomas - 15.11.2024 10:48

thanks to popular demand I put all the music in the description, tell ur friends

Ответить
@TheCoder-o2d
@TheCoder-o2d - 28.03.2025 10:00

how did you managed to import custom ai to tf2

Ответить
@gambit_sfm
@gambit_sfm - 26.03.2025 05:17

I've got no idea if this is a weird thought but if A.I could be used to improve A.I enemies in games like TF2, I think it'd be pretty cool to see a native bot rocket jump for example or see a Spy pull off a tricky stair-stab
just an idea on how I think A.I could actually benefit

Ответить
@jamesmurphy9784
@jamesmurphy9784 - 24.03.2025 12:06

I don’t know if this makes sense I don’t really code and I don’t really know how ai works but couldn’t you slightly reward the bots for staying in the middle of the path cause it seemed like they were always moving to the sides

Ответить
@Ninjayuvibear
@Ninjayuvibear - 24.03.2025 03:41

The jump academy bot

Ответить
@rekardo3
@rekardo3 - 23.03.2025 15:19

:o

Ответить
@graham1616
@graham1616 - 19.03.2025 16:50

Seeing how much tweaking it takes for the bots to work, it makes you think about how impressive the human brain is

Ответить
@javierrc5794
@javierrc5794 - 19.03.2025 14:25

He bout to revive the bot crisis with robo-troldiers

Ответить
@Geomun22
@Geomun22 - 19.03.2025 09:11

Ive never in my life expected these type of ai testing videos to ever FAIL so EPICLY that I CHEERED when this video ended with ai learning programs. You sir, have made me curious again and thatz pretty neat 👌 new sub immediately

Ответить
@subbinbacktoallsubbs5272
@subbinbacktoallsubbs5272 - 18.03.2025 07:41

"What's the fun of showing a baby how things are done?"😂

Ответить
@Onlesi
@Onlesi - 16.03.2025 19:32

Please use red to blue scale in the future for a us colorblind users.

Ответить
@FxekFnf
@FxekFnf - 16.03.2025 02:14

Any chance of using colour blind friendly colours on the spectrums next time (where you put brain time)

Ответить
@IsaiahJaya
@IsaiahJaya - 15.03.2025 15:45

source_author:thomas enabled
videoLikeStatus = True
channelSubscriptionStatus = True
if comment == "Liked";
cookie = n
give source_author:thomas cookie:+1

Ответить
@F2PP
@F2PP - 15.03.2025 04:08

Ngl that is very impressive

Ответить
@ben-frey7864
@ben-frey7864 - 15.03.2025 01:02

What i took away. Be ethical with your penalties so they have freedom too but dont become psychos

Ответить
@filipfurs7902
@filipfurs7902 - 14.03.2025 03:09

This video makes me want to complete jump maps flawlessly

Ответить
@shud713
@shud713 - 13.03.2025 20:19

Didn’t they make a AI TAS like 4-5 years ago for JA2?

Ответить
@ohno6528
@ohno6528 - 10.03.2025 09:21

I see. So AI does worse than humans because they don't have enough data on the enviroment.

Ответить
@SilverDashie
@SilverDashie - 10.03.2025 06:29

Now with the SourceCode release you could probably make some native implementations

Ответить
@Zoneyjah
@Zoneyjah - 08.03.2025 23:09

Bro, this has gotta turn into a mod. Imagine the fun we’ll have with bots if they can pull off player-like moves too?

Ответить
@melatonin1371
@melatonin1371 - 08.03.2025 14:05

First 30 seconds is a bold claim

Ответить
@superkuchen3785
@superkuchen3785 - 08.03.2025 04:28

This is how Dymanic NPS were created

Ответить
@VogelBoys74
@VogelBoys74 - 07.03.2025 20:06

Punish them for failure.
Also are you able to make a A.I beat Half-Life 2 or Portal 1?

Ответить
@Qefix
@Qefix - 07.03.2025 18:16

So that means... I am an AI? 😢

Ответить
@wafflerthe
@wafflerthe - 07.03.2025 14:42

Have you tried using NEAT?

Ответить
@The4Tifier
@The4Tifier - 06.03.2025 09:09

A better way for AI to learn is not to start from absolute scratch, but to use successful human data as their baseline (so that they can refine a successful formula and not have to reinvent it).

Ответить
@dakmanworld
@dakmanworld - 05.03.2025 07:31

shout out to all the actors that posed as bots in the video, truly inspiring

but for real, this was super in-depth and interesting!!

Ответить
@warframewarframe5095
@warframewarframe5095 - 03.03.2025 11:02

Pshh im way better than that robot

Ответить
@klb69
@klb69 - 01.03.2025 16:15

ахуеть

Ответить
@vaan_
@vaan_ - 20.02.2025 14:35

jerma dumpy

Ответить
@LordRazer3
@LordRazer3 - 19.02.2025 10:11

The machine can learn...but a human can learn faster. The machine however can learn perfection more so then man...but if a single action can throw everything off is it really perfection?

Ответить
@Trtko-y2p
@Trtko-y2p - 18.02.2025 00:19

cmon man, make them learn on human acts mimicking them..

Ответить
@lahxialo9894
@lahxialo9894 - 15.02.2025 19:44

Teach an ai to sticky launcher jump

Ответить
@BiteTrend
@BiteTrend - 12.02.2025 09:26

Voice too quiet music and sound fx too loud

Ответить
@Dregomz02
@Dregomz02 - 11.02.2025 12:20

This video from perspective of someone watching it in 2035 and 2045 would be interesting. I will come back there for sure...

Ответить
@gubygamer5522
@gubygamer5522 - 10.02.2025 09:59

lvl 5 the ultimate human power the AI doesn't posses , GIVING UP

Ответить
@anitoancognita7060
@anitoancognita7060 - 09.02.2025 07:42

Bro put mann milk inside his machine💀
Tho i'm curious what if you made the bots to anticipate the long term reward for that pogo jump tile

Ответить
@omartaipe9338
@omartaipe9338 - 31.01.2025 00:25

Maybe try change the reward.
Add a value of reward for the bots while they stay in air or while the dont touch the floor
The objective of that course isnt just reaching the goal, is doing the pogoing too

Ответить
@SomeRandomGuyThatEditsStuff
@SomeRandomGuyThatEditsStuff - 28.01.2025 18:47

"It's learning"

Ответить
@RadarFinsR
@RadarFinsR - 26.01.2025 16:05

"reward" would probably be better analogous to dopamine i.e. feeling good which is how we motivate our own reward centers of our brains.

Ответить
@kitmapking
@kitmapking - 26.01.2025 15:07

Very cool!

Ответить
@3dz3dz
@3dz3dz - 24.01.2025 20:24

Its possible to pogo across that at the fire rate you chose, right?

Ответить