Deepseek V4 explained. #ai #aitools #ainews #llm #agi #deepseek #claude #agi Thanks to our sponsor Abacus AI. Try ChatLLM & DeepAgent today: http://chatllm.abacus.ai/?token=aisearch Deepseek v4: https://api-docs.deepseek.com/news/news260424 LLMs explained: https://youtu.be/U2hZFMVNSE0 Residual connections: https://youtu.be/2IfAVV7ewO0 0:00 Deepseek V4 intro 1:00 Deepseek V4 specs 2:06 The challenge of 1M context 4:16 Hybrid attention 5:11 CSA & sparse selection 6:50 HCA 8:22 Sliding window attention 10:44 Insane efficiency gains 12:02 Signal explosion 13:00 Residual connections 13:52 mHC 14:17 ChatLLM 15:24 mHC continued 17:54 Muon 19:26 Infra challenges 22:31 Training challenges 24:09 Anticipatory routing 25:24 SOTA results 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
🗣️"DEEPSEEK WAS ABLE TO BUILD THIS IN A CAVE!!"
Necessity is the mother of invention as they say.
Deepseek are the goats for releasing their research papers publicly 🥳🎉
I can't believe you broke down the paper this well....what a masterpiece of a video!
deepseek feels like it was actually developed by the people for the people
AI stopped sleeping once Deepseek woke up 🤖
Deepseek is the only company with highly technical CEO.
My guy summarizing better than professors and ai
Deepseek is for the people :D
thanks to China i dont have to pay 25 dollars per 1m token. Competition is a good thing.
if it was done by open ai or any other American owned LLM it would've been all over tech news lmao
Deepseek should sponsor this video. I don't see any better way to explain this to make the layman truly understand the ingenuity here. You're good man
dude i can't wait for 2027 or 2028, i wonder what ai will look like by then
Deepseek is OUR favourite model ❤🤖🐳
Thanks for breaking down such a complex paper to make it easily understood. Nice work sir!
I want whatever the DeepSeek team are smoking.....
Even though Deepseek has open-sourced everything, Anthropic still accuses them of distilling their model. Distillation is a perfectly normal model training technique, no different from how a normal user uses a model, and it has been stigmatized. It's like a competitor copying a restaurant menu – normal business competition. If Anthropic had leaked their model code, do you think they would have included algorithms from Deepseek or other Chinese open-source models?
Deepseek: OUR ai.
My girlfriend keeps telling me "Attention is all you need."
That's a very good explanation! I like it