I always saw this "Least Squares" method on "How to make AI" videos but I never could understand why did we do that. But this video makes it VERY clear :) I feel like I can make chat gpt in 1 day now lol
You have an excellent ability to teach this very complex topics in an entertaining fashion. The animations also make it very easy to follow. Thank you for all your work. ❤❤
small correction, gradients tell which way function increase, we take negative of gradients to decent (but not always eg rl where you want to maximize reward)
maybe the obstacle in artificial intelligence to be able to think like human is the capability to memorize and think at the same time? like for us, our memory can be like a "time machine" that we can go back to, and with our capability to rationalize, imaginate and think foward makes us "smart". something that a robot with zeros and ones can't
Hm? Being based on 1s & 0s doesn’t prevent retrieving records of previous inputs and processing. And, whatever processing can be done, can be done in relation to those retrieved records.
@@vinniepeterss Well, you can have e.g. a video feed, and a program can do processing on recorded video, right? (Doesn’t have to be video, just an example) Maybe I’ve misunderstood what you meant. Did you maybe mean something like, being able to efficiently find the records of past events that are relevant to what is currently going on? I think that task is probably one which people might be significantly better at than computers, and maybe in a way that results from how the computer hardware works? Idk. That seems at least sorta similar to what you were suggesting?
@@drdca8263 well, the way i look at it is, that computer cannot "think" back and foward like the way we humans do it i guess? ok. for example if i need you to answer some idea that never been tested before and we want to test it, the way we do it is to remembering what we learn with respect to the idea itself (memorizing, but sorta looking back) and combine with speculative results (imaginatively). that's what the computer cannot do maybe? i dunno man, i come from third world country. maybe when i wrote that comments i'm full of thoughts but now i'm lost jajajaja
because doing it for say, 20 data points vs 20 million data points requires a large amount of extra knowledge, and starts introducing engineering problems, not just mathematical ones.
This has to be the best Intro to "Calculus in ML" explanation video ever. Hats off!!!
This video MUST have 10s of millions of views
I always saw this "Least Squares" method on "How to make AI" videos but I never could understand why did we do that. But this video makes it VERY clear :) I feel like I can make chat gpt in 1 day now lol
Hands down the best explanation of the chain rule I've ever seen!
How does this channel not have more subs? Every video is an object lesson in clarity.
Amazing video. Really enjoyed and learned a lot. Thanks!
You have an excellent ability to teach this very complex topics in an entertaining fashion. The animations also make it very easy to follow. Thank you for all your work. ❤❤
I love this series so far!
that was awesome! looking forward to part 3!
Wow If I take your math class, I could join MIT, Stanford or Oxford! high Quality stuff!
OMG. It's so clear now. Thank you!
This is really well done. Thank you!
awesome summary! keep it up!
Thanks a lot, keep the GREAT work
I love your videos, Brian.
small correction, gradients tell which way function increase, we take negative of gradients to decent
(but not always eg rl where you want to maximize reward)
this was beautiful
fantastic
beautiful. THank you brian
This has unlocked AI for me. Thanks!
maybe the obstacle in artificial intelligence to be able to think like human is the capability to memorize and think at the same time? like for us, our memory can be like a "time machine" that we can go back to, and with our capability to rationalize, imaginate and think foward makes us "smart". something that a robot with zeros and ones can't
Hm? Being based on 1s & 0s doesn’t prevent retrieving records of previous inputs and processing.
And, whatever processing can be done, can be done in relation to those retrieved records.
@drdca8263 oh, really? i didn't know that. i'm just speculating...
@@vinniepeterss Well, you can have e.g. a video feed, and a program can do processing on recorded video, right? (Doesn’t have to be video, just an example)
Maybe I’ve misunderstood what you meant. Did you maybe mean something like, being able to efficiently find the records of past events that are relevant to what is currently going on?
I think that task is probably one which people might be significantly better at than computers, and maybe in a way that results from how the computer hardware works? Idk.
That seems at least sorta similar to what you were suggesting?
@@drdca8263 well, the way i look at it is, that computer cannot "think" back and foward like the way we humans do it i guess? ok. for example if i need you to answer some idea that never been tested before and we want to test it, the way we do it is to remembering what we learn with respect to the idea itself (memorizing, but sorta looking back) and combine with speculative results (imaginatively). that's what the computer cannot do maybe?
i dunno man, i come from third world country. maybe when i wrote that comments i'm full of thoughts but now i'm lost jajajaja
When you described how gradient descent works the first thing I thought was "what about local minima?".
Anyway, thanks for the video!
great video.
Can you cover modules?
❤❤
nice
Let the machine cooks 🗣️🔥🔥
Where's the third part 😭
Backpropagation?
ok
Holy shit es war echt wii sports, ich habs gedacht aber dachte das kann eigentlich nicht :D
tbh I'm not interested in ML. I hope you get back to less scab/scam based nicer algorithms.
Not everything that you don't understand is a "scam".
Thats just statistics thats it. Idk why implementing this on a computer gives it another name
Numerical math*
because doing it for say, 20 data points vs 20 million data points requires a large amount of extra knowledge, and starts introducing engineering problems, not just mathematical ones.
you can say everything is mathamatics in that way, why have names for anything.
@@ooogabooga5111 wdym, they already have names. im talking about renaming
mathematics is the art of giving different names to the same things
Next video when?