The lesson oligarchs should learn from DeepSeek is that sharing is caring. To illustrate my point, here are some butthurt cream.
Miscellaneous News
- Thread starter bd popeye
- Start date
The lesson oligarchs should learn from DeepSeek is that sharing is caring. To illustrate my point, here are some butthurt cream.
Not actually impressed-just cowering/hating as is their habit but will stab you in back at first opportunity(no machine tools/precision machines for them China)
Monolithic vs Microservice
I am not talking about websites, even though the concept of microservice started from web application architecture. I am talking about methodologies in general. I am not boasting here - I am an expert in microservice architecture, in cloud or on-prem. Maybe some people would argue microservice and on-prem are oxymoron. But in my practices, software architecture has become ever more platform agnostic.
So why is above related to DeepSeek vs NIM?
From software architecture standpoint, up until DeepSeek R1 release, all hyperscalers deploy LLM models in a monolithic architecture. History of software engineering has already demonstrated that paradigm is shifting toward microservice architecture from monolithic architecture, particularly loosely-coupled micro-service architecture. There are exceptions in corner cases in various vertical problem domains. But general trend for large-scale cloud-native SaaS is going loosely-coupled micro-service architecture. DeepSeek R1 is the first top-performing tier-one model that can fit in, while other top models that hyperscalers have deployed are all monolithic that can only be deployed in mega data centers with humongous GPGPU farms. Therefore, one of the fundamental impacts DeepSeek brings upon the current AI revolution is that it has started a paradigm shift in deployment architectures. In other words, it is not that DeepSeek is cheap and free that bring so much fear, rather it is innovations in its implementation techniques as well as overall software architecture that has fundamentally change the game of generative AI.
This is just the beginning ......
Guys, Rubio knows it now!
Seems like the translator he listend to during the call is from Chinas side, and he/she didn't done his job well. LOL
At least this one didn’t have blue hair.
keep this CIA bitch OUT!!!!!! I mean it China/MSS(full security for the team) this deepseek episode is no joke to these westoid losers of 2-3 trillion dollars/massive global tech hit -who the hell would hop on plane to Liang's home village -while speaking perfect Mandarin??"Where is your boy grandma,where???"
manqiangrexue
Brigadier
Yeah, now they're just waiting for us to commit some kind of genocide before, "Sar, we are elite fighter, like Nazi SS!" turns into, "Sar, we are elite fighter like Chini brother PLA death squad!"
You know, FYI there are a lot of fanboys exist in China too.
yeah sure one hotel can bribe him. unless you understand these things. I will not switch to Arabic what that really means.
If DeepSeek's hedge fund owners shorted NVIDIA, they made more profit than OpenAI in its entire existence.
Because OpenAI has never been profitable, ever, and is getting more unprofitable by the second.
And now it never will be without massive government subsidies. All the loss is now locked in.
If DeepSeek did not go open source, OpenAI still has a possibility of being profitable with a captive market. They'll just ban DeepSeek. It then wouldn't matter how cheap it is. It wouldn't matter if it was even free. Tik Tok is free and got banned too. Open source is the only way to actually make a difference. You can keep a superior or customized model close source if necessary.