It feels good to have you back Siraj Raval. I mean the REAL you. the you that talks about stuff in AI that really matters and dearing to get into the deep deep R&D take on AI. I always consider you as one of the original visionaries and pioneers in teaching and promoting and leading enthusiast and professional forward with inspiration and ideals. Good to have you back hope the stigma and condemnation of the past stays in the past and ur up there like other AI channels like "yannic kilcher (by Yannic Kilcher)" "machine learning street talk (w/ Tim Scarfe and Keith Duggar)" and so many others. Never forget you are one earliest earliest originals.
Interesting conjecture! I've been chasing down a lot of the same papers/have the same pet project, will have to take a closer look at your implementation
evaluations and benchmarks missing. The current evaluation is a simple check of the model's performance on arithmetic problems. It generates a batch of arithmetic problems, runs them through the model, and computes an average reward based on whether the model's output matches the expected result. How is it compared to otther o1 models, gpt4o, claude sonnet etc?
I’m sorry to be “that negative guy” in the comments, but some of your claims here are overstretched, and the concepts you’re throwing around are at a surface level. You made little reference to the importance of reward models in PPO and did not distinguish between per-step and global evaluation (a critical aspect of creating the tree structure you made reference to). There’s also no evidence that reasoning models require special tokens. Finally, the applicability of your method here is super constrained, whereas other MCTS-based methods with language models manage to generalize to non-math based tasks. You’ve produced excellent videos in the past, but this one unfortunately falls short
Thanks for the feedback, I'll include those missing points in the next video on o1. There are a ton of details i missed out on that i'll add to the next one.
seeing the thumbnail i thought you provided some code to add o1 nano to chatgpt web app, but ended up knowing it is something which needs to be run in vs code or something. :( , you might be technically strong, but dont create excitement with thumbnails like that, if you show soemthing in thumbnail then please create and provide us what you showed in thumbnail
But why? How is this video benefiting anyone ? Ntng learned , ntng interesting. I used to like ur videos , but unfortunately ur content are not exciting anymore
We need more videos like this! Also, it would be great if you started by fine-tuning existing models on chains of thought (PRM800K) for reasoning and demonstrating how reasoning evaluations work. You can even do that on OpenAI Playground to make it beginner-friendly. Imagine how many people will start experimenting with their own datasets!🎉
@@SirajRaval😂 I know right. The world is now full of people with low attention spans. They want an easy way to put money in their pockets. Most people are not as smart as you! Those fast money sports betting colabs or some cplex algorithm based crypto or sports betting arbitrage colab notebooks is all you need to put out to keep people engaged. Trying to recreate O1 models is for 0.01% of those who watch your channel. You’ve been doing a great job tho. I’m also surprised you are not behind any music AI companies. Maybe you are! You were super early on the music AI wave. Waay before the Sunos and Udios. I wish you well my brother
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
Havent seen or kept up with your channel in a long while but glad to see you're still creating awesome and well-explained content!
I'm feeling extra passionate about democratizing AI lately the more closed OpenAI gets, welcome back!
Ha this is great, I was thinking a few days ago about what would happen if we used o1 to document itself and the paper chain, and you went and did it!
i have been waiting for you to make AI tech videos again from so long.
your videos were my introduction to AI domain ,Thanks
Give this man the compute he will give you opensource O1
Awesome man, great to see you back into AI.
Awesome to see you again, continue the great work
Doing great job. Keep going!!!.
Missed you all along these years.
It feels good to have you back Siraj Raval. I mean the REAL you. the you that talks about stuff in AI that really matters and dearing to get into the deep deep R&D take on AI. I always consider you as one of the original visionaries and pioneers in teaching and promoting and leading enthusiast and professional forward with inspiration and ideals. Good to have you back hope the stigma and condemnation of the past stays in the past and ur up there like other AI channels like "yannic kilcher (by Yannic Kilcher)" "machine learning street talk (w/ Tim Scarfe and Keith Duggar)" and so many others. Never forget you are one earliest earliest originals.
aitutorialmaker AI fixes this. Siraj Raval explains ChatGPT O1.
Siraj rocks ..
back to track great
Love your videos! 🥰
Back again!! Keep publishing this type of videos for which you were known for!! 😊
One question that I have is that, which dataset you have train your model?
Can you make it available.
Where are you, we need you. I jumped into this side of the world because of you
Interesting conjecture! I've been chasing down a lot of the same papers/have the same pet project, will have to take a closer look at your implementation
evaluations and benchmarks missing. The current evaluation is a simple check of the model's performance on arithmetic problems. It generates a batch of arithmetic problems, runs them through the model, and computes an average reward based on whether the model's output matches the expected result. How is it compared to otther o1 models, gpt4o, claude sonnet etc?
Great points. It needs a more in depth analysis, I’ll do that next vid
Few people teach AI as well as Siraj,❤😊
mind blowing
I’m sorry to be “that negative guy” in the comments, but some of your claims here are overstretched, and the concepts you’re throwing around are at a surface level. You made little reference to the importance of reward models in PPO and did not distinguish between per-step and global evaluation (a critical aspect of creating the tree structure you made reference to). There’s also no evidence that reasoning models require special tokens. Finally, the applicability of your method here is super constrained, whereas other MCTS-based methods with language models manage to generalize to non-math based tasks.
You’ve produced excellent videos in the past, but this one unfortunately falls short
Don't even bother, he thinks he's right about everything.
Thanks for the feedback, I'll include those missing points in the next video on o1. There are a ton of details i missed out on that i'll add to the next one.
Thanks Siraj
Nice !
awakening is real
I can't signup to tradergpt, error
Cool
Ilya Sutskever was NOT one of the authors of "Attention is all you need".
you are absolutely right. thanks for pointing that out. my mistake
nice hair cut ;)
What's the ideal strategy to take advantage
of the current crypto bull market
As a novice, it's vital to have a mnentor
for accountability. I'm being advice by a
reputable crypto consultant.
She goes by Lisa
Oh, do you happen to know her too? I'm
proud to say I've also benefited from her
platform
Archieving a weekly profit of 45k
I used to think I was the only one
benefiting from her guidance in
navigating the volatility of trading.
It doesn't even use the pre-trained model. I don't think it works
Cheater.
seeing the thumbnail i thought you provided some code to add o1 nano to chatgpt web app, but ended up knowing it is something which needs to be run in vs code or something. :( , you might be technically strong, but dont create excitement with thumbnails like that, if you show soemthing in thumbnail then please create and provide us what you showed in thumbnail
will do. thanks
o1 not O1 !!!
In an investigation, details matter - Jack Reacher
But why? How is this video benefiting anyone ? Ntng learned , ntng interesting. I used to like ur videos , but unfortunately ur content are not exciting anymore
To democratize AI knowledge, just like I did before. This video is pretty exciting, sad you lost your enthusiasm after all these years
We need more videos like this! Also, it would be great if you started by fine-tuning existing models on chains of thought (PRM800K) for reasoning and demonstrating how reasoning evaluations work. You can even do that on OpenAI Playground to make it beginner-friendly. Imagine how many people will start experimenting with their own datasets!🎉
@@SirajRaval😂 I know right. The world is now full of people with low attention spans. They want an easy way to put money in their pockets. Most people are not as smart as you! Those fast money sports betting colabs or some cplex algorithm based crypto or sports betting arbitrage colab notebooks is all you need to put out to keep people engaged. Trying to recreate O1 models is for 0.01% of those who watch your channel. You’ve been doing a great job tho. I’m also surprised you are not behind any music AI companies. Maybe you are! You were super early on the music AI wave. Waay before the Sunos and Udios. I wish you well my brother