I think the community is fine to publish this content, as long as it is about technology.
I found it super interesting how you managed to train the AI to play google dinosaur better haha. Someday I will delve into the topic of AI's, I am still quite inexperienced in the subject.
Talking about AI, it just blew up in recent years. I always knew it is going to be a big thing, I and a friend started learning it together. He continued on and now became a good machine learning engineer in this field. I didn't want to go in that direction even after knowing its potential. Call me stupid but I play by my rules.
(If I had made that decision, I might have been much better financially but perhaps a lot worse in general. So, I chose a different path and I am loving this, no regrets...lol)
The important thing is that you did what you felt haha. The economic part is not always everything.
Maybe someday I'll pay for some training to learn more about AI. I had read that you used deepseek. I tried one of their models on a VPS and it really wasn't that good, I think it still has some way to go to beat openAI.
R1 has multiple deployable model ranging from 1.5 billion parameters (weak) which even I can run on my system to 671b parameters model (needs 32 GB VRAM GPU and ~400 GB Storage). This one is the strongest, but takes a lot more resources to deploy. We just got a gaming GPU with that much VRAM 5090. $2K for a GPU is insane though 🤪
It's an interesting experiment man, I wouldn't have thought that the AI would have such a hard time playing this game because it's so easy for a human. And if you were to run this experiment for 1-3 days, do you think it could be perfected?
Ideally, the more training it undergoes, the more it learns. But to perfect it I might have to feed it more information like its current distance from the nearest obstacle, maybe the number of obstacles in the scene, its vertical height, maybe a parameter for the changing game speed etc.
That will make it observe a lot of things every step and will improve it a lot. But as it gets complex, it needs more calculations per step and needs much more computational power, which my system doesn't have. Also, all the past experiences have to be stored in memory and for a large model you might need tens of GB of free RAM or VRAM if not 100s of GB.
But in my current state, I can train it for a few more hours and see if it learns more. But I stopped the training because it was getting slow. Also, if it wasn't a web game, you can run 100s of games in parallel and the training time will be much shorter. But again, you need more processing power to run all of those simultaneously.
I think the community is fine to publish this content, as long as it is about technology.
I found it super interesting how you managed to train the AI to play google dinosaur better haha. Someday I will delve into the topic of AI's, I am still quite inexperienced in the subject.
That's what I thought.
Talking about AI, it just blew up in recent years. I always knew it is going to be a big thing, I and a friend started learning it together. He continued on and now became a good machine learning engineer in this field. I didn't want to go in that direction even after knowing its potential. Call me stupid but I play by my rules.
(If I had made that decision, I might have been much better financially but perhaps a lot worse in general. So, I chose a different path and I am loving this, no regrets...lol)
The important thing is that you did what you felt haha. The economic part is not always everything.
Maybe someday I'll pay for some training to learn more about AI. I had read that you used deepseek. I tried one of their models on a VPS and it really wasn't that good, I think it still has some way to go to beat openAI.
Interesting...which model did you use exactly, there are some distilled models too. Was it the larger 671 billions parameter model or not?
Deepseek R1 i think
R1 has multiple deployable model ranging from 1.5 billion parameters (weak) which even I can run on my system to 671b parameters model (needs 32 GB VRAM GPU and ~400 GB Storage). This one is the strongest, but takes a lot more resources to deploy. We just got a gaming GPU with that much VRAM 5090. $2K for a GPU is insane though 🤪
https://ollama.com/library/deepseek-r1:671b
We did it on a VPS with low specs, but it took too long to develop a response and gave false information. Perhaps it is because of the components
It's an interesting experiment man, I wouldn't have thought that the AI would have such a hard time playing this game because it's so easy for a human. And if you were to run this experiment for 1-3 days, do you think it could be perfected?
Ideally, the more training it undergoes, the more it learns. But to perfect it I might have to feed it more information like its current distance from the nearest obstacle, maybe the number of obstacles in the scene, its vertical height, maybe a parameter for the changing game speed etc.
That will make it observe a lot of things every step and will improve it a lot. But as it gets complex, it needs more calculations per step and needs much more computational power, which my system doesn't have. Also, all the past experiences have to be stored in memory and for a large model you might need tens of GB of free RAM or VRAM if not 100s of GB.
But in my current state, I can train it for a few more hours and see if it learns more. But I stopped the training because it was getting slow. Also, if it wasn't a web game, you can run 100s of games in parallel and the training time will be much shorter. But again, you need more processing power to run all of those simultaneously.
That's something interesting now... But beware to never reach the ending as I already have seen it.
Tiktok video
This model can't, it is very basic and runs sequentially so it won't reach anywhere close to the end.