How GPT and other large language models (LLMs) work. Transformers explained. How AI works. #ai #llm #machinelearning #datascience #transformer Thanks to our sponsor Genspark. Try it for free https://bit.ly/4uM3PLS Attention is all you need https://arxiv.org/html/1706.03762v7 0:00 Intro 0:33 The transformer model 1:30 Predicting the next word 2:30 Tokenization 5:06 Representing meaning 7:17 Positional encoding 9:17 Attention head 14:49 Genspark 16:35 Multiple heads 19:30 Add and norm 21:45 Feed forward neural net 24:08 Multiple decoder blocks 24:50 Final layer 27:03 Training the model Newsletter: https://aisearch.substack.com/ Find AI tools & jobs: https://ai-search.io/ Support: https://ko-fi.com/aisearch Here's my equipment, in case you're wondering: Lenovo Thinkbook: https://amzn.to/4jWeKwH Dell Precision 5690: https://www.dell.com/en-us/dt/ai-technologies/index.htm?utm_source=AISearchTools&utm_medium=youtube&utm_campaign=precisionai#tab0=0 GPU: Nvidia RTX 5000 Ada https://nvda.ws/3zfqGqS Mic: Shure SM7B https://amzn.to/3DErjt1 Audio interface: Scarlett Solo https://amzn.to/3qELMeu
ADVERTISEMENT
I’ve spent a lot of time digging into how these models actually work and this is one of the rare explanations that makes the whole next-token thing finally click!
Man, this is like the best video explaining transformers ever made .
Brother..I'm a music composer from India and I've been following ur channel for months now...I actually maintain notes which I write based on ur weekly ai updates...u have been an angel in my ai journey..thanks a lot
AI news so dry right now bro made a explainer, its a good one tho!
As someone who recently studied the LLM architecture i highly recommend this video for anyone trying to understand the basics. Beginning and end are phenomenal explanations, the minimalistic visuals fit perfectly for the topic and are absolutely on point. The Add & Norm Phase could have been a little bit more detailed, thats the only part which could bear confusion for someone not familiar with the system, but since its already 30 minutes of dense information taking the cut wasnt a bad decision. Big thumbs up, you have by far the best AI-related news channel and even if instructional videos should adress less viewers: Having a solid amount of these similar learning material opens the possibility to be far more than the original purpose of a news channel.
I quite like this "AI Explained" format, very well explained and easy to follow. Please do Diffusion Transformers next!
This is by far the best video that explained this easily like spoon feeding..
Is still a mystery to me why the authors of the original paper attention Is All You need haven't received the Nobel Prize yet. It's hard to think of any single paper that has had a greater impact on society and human existence.
Who even came up with these math methods that can capture our human logic to understand context? Such a genius
Thanks to our sponsor Genspark. Try it for free /4uM3PLS
This is the best explanation so far! Thank you
this is genuinely the best an clearest explanation video of transformers I've ever watched. massive props to you
The amt of times i got questions asked from this paper is just insane😭😭 during my college days
Even after watching this explanation, it still feels like magic to me
I literally was researching this up this morning
Love using this channel to learn about AI simply
This video made everything super simple and easy to learn. I had a basic idea of how transformer models worked but now I feel like I truly understand them now. W video
Shoutout to AlexNet, which I believe is the reason "Attention Is All You Need" exists.
Now this is going to be interesting, thank you.
Wow this part of the video explains everything with just the student on red, reading 😂😂😂 30:26