You do know this of course, bug perhaps to put this into context for others. Nvidia deploys a bunch of ML Models actually, many of them top of their class. Thing is it's mostly not LLMs, and the one they apparently do distribute (ChatRTX) is made to run on one (!) of their cards. It's a software add-on, like the many other add-ons they create (eg. DLSS, RTX Voice, also ML add-ons). They are supposed to help the sale of cards, and I'd wager they're also a nice way for Nvidia to dogfood their own software solutions (at least during initial phases). The problem is that the one LLM they ship that way can't possibly compete with the "superlarge" - as I like to call them - models that run in the cloud. For starters Nvidia is stingy with their VRAM on consumer cards (being able to segment away more profitable business sales probably isn't innocent in that), but even if they would double or triple the VRAM, it'd still be basically impossible to beat. And its also not really possible to protect such an investment. OpenAI can monitor API usages and at least try to cut off companies that use it to copy GPTX, though evidently - grok - they don't do that perfectly. But Nvidia would ship the model parameters to every consumer, no way that can't be reverse engineered. They would have to do it as a service, just like every other AI company does these days. Nvidia isn't foreign to this, they've got... That remote gaming PC service, whatever they call it again. But it's not really their core competence. But perhaps more importantly, it would put then into direct competition with their company customers. And if there's one thing companies HATE, it's their supplier competing with them. Unless there is a huge power misbalance (which there isn't between Nvidia and Microsoft etc) this usually doesn't end well for the relationship between supplier and customer. If Microsoft and friends are looking for alternatives right now, this would lead to a multiplication of that effort. Nvidia has a good thing going there, while they may very mich be tempted, I don't think they will ever commit enough resources to something like that, unless perhaps they've got a real big advantage nobody can copy. It's already extremely risky to compete in the big league, if that has the potential to destroy your core business it gets exponentially more risky.
Sir i live in too much noisy place does foam reduce noise enough that i play in 100 vol they dont listen as they ply 100 vol I don't listen/or any other solution you know
U must be new to cryptocurrency or not paying attention its full of scams and rug pulls . Money lost is so hight that number of people seeking regulation is growing and without Central authority it cant work EVER
i just randomly came across this video and not expecting such great knowledge ..... just one thing excellent work sir
So nice of you
Your channel is worth bing watching specially the one's on unseen societal issues
do u mean "social " anyway Thanks for the visit
A lot of valuable info, thanks for video!
Glad it was helpful!
You do know this of course, bug perhaps to put this into context for others. Nvidia deploys a bunch of ML Models actually, many of them top of their class. Thing is it's mostly not LLMs, and the one they apparently do distribute (ChatRTX) is made to run on one (!) of their cards. It's a software add-on, like the many other add-ons they create (eg. DLSS, RTX Voice, also ML add-ons). They are supposed to help the sale of cards, and I'd wager they're also a nice way for Nvidia to dogfood their own software solutions (at least during initial phases).
The problem is that the one LLM they ship that way can't possibly compete with the "superlarge" - as I like to call them - models that run in the cloud. For starters Nvidia is stingy with their VRAM on consumer cards (being able to segment away more profitable business sales probably isn't innocent in that), but even if they would double or triple the VRAM, it'd still be basically impossible to beat. And its also not really possible to protect such an investment. OpenAI can monitor API usages and at least try to cut off companies that use it to copy GPTX, though evidently - grok - they don't do that perfectly. But Nvidia would ship the model parameters to every consumer, no way that can't be reverse engineered.
They would have to do it as a service, just like every other AI company does these days. Nvidia isn't foreign to this, they've got... That remote gaming PC service, whatever they call it again. But it's not really their core competence.
But perhaps more importantly, it would put then into direct competition with their company customers. And if there's one thing companies HATE, it's their supplier competing with them. Unless there is a huge power misbalance (which there isn't between Nvidia and Microsoft etc) this usually doesn't end well for the relationship between supplier and customer. If Microsoft and friends are looking for alternatives right now, this would lead to a multiplication of that effort. Nvidia has a good thing going there, while they may very mich be tempted, I don't think they will ever commit enough resources to something like that, unless perhaps they've got a real big advantage nobody can copy. It's already extremely risky to compete in the big league, if that has the potential to destroy your core business it gets exponentially more risky.
Well put
I think that AI is somehow similar to Industrial revolution of 1700s
Maybe
What to do, then pick up Nvidia's share 😊
I dont know where we are at the curve
@@s2tenglish lets try our luck
Sir i live in too much noisy place does foam reduce noise enough that i play in 100 vol they dont listen as they ply 100 vol I don't listen/or any other solution you know
put mic as close to you as possible andFoam Mattress is cheaper & Bigger
According to your thoughts when could nvidea get boom
Are you ghey ?
its wont be a BOOM per say more like DIP
What's your thought on cryptocurrency can it makes decentralised banking and transactions possible
U must be new to cryptocurrency or not paying attention its full of scams and rug pulls . Money lost is so hight that number of people seeking regulation is growing and without Central authority it cant work EVER
@@s2tenglish Yeah I am new
It's time to go global. Nice try sir.
Let's see