Little Learning Machines is available right now on Steam, with a free update coming September 5th: store.steampowered.com/app/1993710/Little_Learning_Machines/ Find out more about Little Learning Machines via the links below. Website: www.transforms.ai/little-learning-machines Learning Companion (PDF): www.transforms.ai/learning GDC Postmortem Talk: gdcvault.com/play/1034597/AI-Summit-Little-Learning-Machines Plus catch my bonus conversation with Dante, Nick, and Fiona over on the Branching Factor podcast on AI and Games Plus (plus Spotify, Apple Podcasts and more!) kzitem.info/news/bejne/0X1tk6mYm4CLeX4 -- Thanks again to the team for having me work with them on this project - they were great to chat with. I hope you all enjoyed learning about this quirky wee game.
@aedan_s
12 күн бұрын
Rest in peace Anuj Patel 🙏You helped bring this game to life and helped pioneer new game ML at runtime. It's a shame his passing was glossed over in this video.
@Ixarus6713
13 күн бұрын
This is how game development should be taught! In heavy cooperation with actual games. Imagine a version of Dark Souls that's 10x harder, but with completely open code. Students could open it up and fix its balance, e.g: starting by simply modifying values and eventually even reworking enemy ai and making attacks more delayed. Games are a much better medium to encourage experimentation in coding, especially with people who want to develop games. Good devs! Keep it up!
@notnoodle2196
13 күн бұрын
Looks like a very interesting game.
11 күн бұрын
Eh... not really. They're not that in depth.
@damienkarkof4741
13 күн бұрын
Very cool use of reinforcement learning ! There is so much more things that we could do with reinforcement learning, I hope that more people will try to develop games around it
11 күн бұрын
They did... but it has it's flaws.
@damienkarkof4741
11 күн бұрын
Of course but it is still an open research subject that is still looking for problems to solve and ways to use it. So it would be interesting to see more of it in the game industry
11 күн бұрын
@damienkarkof4741 Indeed. Kind of why I've played it. I have done game development myself since 2018 and started studying ai/ml/rl back in like 2021... coming up with good ideas and use cases is miserable. Also it's rarely close to as efficient as known solutions making it even harder to use.
@damienkarkof4741
11 күн бұрын
I agree that right now that deep learning(LLM or reinforcement learning ) methods are a solution to a non existing problem. For myself, I work mainly in Deep learning and reinforcement learning since 2021. Since then, I have seen learned what problems these are good and the ones that are not. The gaming industry is a really hard case to integrate them into as a feature. They work great to help as a tool but there are not reliant enough to be a core gameplay element. If I look at reinforcement learning, the main application I see is to improve the AI/Behavior of NPC. For this, I think this is advancing on both sides in a good way. In reinforcement learning, we try to focus more on learning human like behavior and adaptation than just on the best performances which is more interesting if you want to feel like you play with or along side a "logical AI". On the other hand, we are trying new approach on AI(without deep learning I mean) in games and we want the player to interact with it differently. I think that we'll have to wait a bit but the intersection between both worlds will happen and will be very cool to experience
11 күн бұрын
@damienkarkof4741 Also humans like systems they can abuse and that are predictable in general..
@remie123
13 күн бұрын
Sounds really cool!
@jamesforsythe3566
13 күн бұрын
This would be a cool game to play in VR
@El-Burrito
13 күн бұрын
while(true); learn! Is also another game that tries to teach machine learning concepts!
@AIandGames
13 күн бұрын
Indeed, it's worth checking out too.
@OzoneTheLynx
13 күн бұрын
Never heard of it, but now I'll have to check this out. Love it!
@kristoferkrus
8 күн бұрын
This is so interesting! I didn't know that it was possible to train something to come up with an intelligent behavior using reinforcement learning that quickly, especially on consumer hardware. Do you know if the developers have published some paper or written some blog post about detailing the method that they used? I would love to read and learn about this work specifically. I did find a paper called "Little Learning Machines: Real-Time Deep Reinforcement Learning as a Casual Creativity Game," but it doesn't go very deeply into the technical details. On the other hand, I also found their session "AI Summit: 'Little Learning Machines' Postmortem: A Game About Training Neural Networks" at GDCVault (I don't know from when), which does seems to be a bit more detailed.
@AIandGames
7 күн бұрын
The most detailed thing they've published to my knowledge is the GDC talk. That was presented earlier this year after the game came out in full.
@diplomadogerontologiasocia1898
13 күн бұрын
Felicidades Dante y su compañero 👍
@El-Burrito
13 күн бұрын
This looks like an interesting game, I want to get it!
@AntonAdelson
13 күн бұрын
14:05 😳😳😳😳😳 i had my fiancee pass away just a month ago. I firmly believe it's important to remember people who die before their time!!! Is there more info on him???
@-C69-
13 күн бұрын
So, how exactly did they get the ML editor in the game? Explain it like a Computer Scientist would… I want to put scratch into Unity… and don’t know if it’s possible (without rewriting the entire scratch code base in c#)
@user-sl6gn1ss8p
13 күн бұрын
Did they? My understanding was that they ended up interfacing with python (hence the needed installation to set up pytorch)
Пікірлер: 27