Google's TurboQuant Is Way Too Overhyped
Related videos
A new way to fine-tune LLMs just dropped
bycloud
16.3k views
Is the One Piece Meal Overhyped
penguinz0
698.2k views
Google's TurboQuant Crashed the AI Chip Market
Wes Roth
57.5k views
The Death of RAG?
bycloud
15.0k views
if you’re overwhelmed by AI tools, watch this
Jeff Su
34.2k views
Is Google expecting too much of opensource
ThePrimeTime
63.9k views
Google's new AI project is UNREAL
Wes Roth
53.5k views
Is It EVEN Possible To Reverse Engineer AI’s Training Data?
bycloud
39.5k views
Google is about to bust the AI bubble...
Wes Roth
77.7k views
I TRADED GOOGLE FOR AI- and you should too
Ian Carroll
238.2k views
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
A new way to fine-tune LLMs just dropped
bycloud
16.3k views
Is the One Piece Meal Overhyped
penguinz0
698.2k views
Google's TurboQuant Crashed the AI Chip Market
Wes Roth
57.5k views
The Death of RAG?
bycloud
15.0k views
if you’re overwhelmed by AI tools, watch this
Jeff Su
34.2k views
Is Google expecting too much of opensource
ThePrimeTime
63.9k views
Google's new AI project is UNREAL
Wes Roth
53.5k views
Is It EVEN Possible To Reverse Engineer AI’s Training Data?
bycloud
39.5k views
Google is about to bust the AI bubble...
Wes Roth
77.7k views
I TRADED GOOGLE FOR AI- and you should too
Ian Carroll
238.2k views
Top Comments (10)
Shh... Let it be over hyped so that RAM prices keep decreasing 😭
Thank you for taking the time to actually understand the paper and explain it to us mortals instead of rushing it right after the release.
Overhyped yes, dishonestly presented absolutely, however, It's still great.
Cant wait for rotorQuant
And remember, you need a hardware capable to handle fp4 or use a black magic custom made for your hardware and pray to not backfire
We are leaving the economy of intelligence and going towards idk what.
"By quantizing our model and diving its precision by 8, we saw a 8x size and performance uplift" Thank you Google for inventing quantization and saving the AI space. You should get a noble price for your creativity and ingenuity.
I saw way more hype coming from people who didn't digest the paper or reacted to someone else's summary that was likely 3 steps removed from the paper than coming from the authors themselves. This is especially true for things like confusing kv cache for total model memory. Same for some of the speed up metrics - engineers will often isolate a component and talk about that without the whole system. For instance a paper might talk about a 10x speedup in a component that is only 5% of the time in real world systems. It's still a huge impact, but doesn't mean the whole system is 90% faster.
thanks for actually presenting the technology and not just mindlessly hyping shit up. i wasnt even aware of all the controversy surrounding it. thanks a lot for showing it to us.
Check out Inngest and let your AI agents wear a harness now! https://www.inngest.com/?utm_source=youtube&utm_medium=video&utm_campaign=yt-bycl-4
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
Shh... Let it be over hyped so that RAM prices keep decreasing 😭
Thank you for taking the time to actually understand the paper and explain it to us mortals instead of rushing it right after the release.
Overhyped yes, dishonestly presented absolutely, however, It's still great.
Cant wait for rotorQuant
And remember, you need a hardware capable to handle fp4 or use a black magic custom made for your hardware and pray to not backfire
We are leaving the economy of intelligence and going towards idk what.
"By quantizing our model and diving its precision by 8, we saw a 8x size and performance uplift" Thank you Google for inventing quantization and saving the AI space. You should get a noble price for your creativity and ingenuity.
I saw way more hype coming from people who didn't digest the paper or reacted to someone else's summary that was likely 3 steps removed from the paper than coming from the authors themselves. This is especially true for things like confusing kv cache for total model memory. Same for some of the speed up metrics - engineers will often isolate a component and talk about that without the whole system. For instance a paper might talk about a 10x speedup in a component that is only 5% of the time in real world systems. It's still a huge impact, but doesn't mean the whole system is 90% faster.
thanks for actually presenting the technology and not just mindlessly hyping shit up. i wasnt even aware of all the controversy surrounding it. thanks a lot for showing it to us.
Check out Inngest and let your AI agents wear a harness now! https://www.inngest.com/?utm_source=youtube&utm_medium=video&utm_campaign=yt-bycl-4