Chinese Technology Continues To Improve And American Tech Should Be Concerned

It's open source because they want max participation to collect more IP. That's Chinas model, let the US and west do the heavy lifting, steal it and make it cheaper so they gain market share. The problem with that is massive and will likely end in war of we don't get this under control. Name the last new product or innovation of come out of China other than COVID? No one can because it all starts over here.They don't have the same protections as Western democracies and unless that changes we will likely see our power, influence and quality of life dwindle.

 

Holding my Nvidia. Added to Broadcom today as it was punished more than NVDA and its half software. This was algos and hedge funds; not retail unwinding this. Also, DeepSeek is owned by a Chinese hedge fund. No coincidence this comes out when Nvidia goes into its silent period ahead of earnings.

 

To clarify, the Deepseek model that caused the selloff was one Large Language Model that did math problems. It was on par with OpenAI and used less datasets to train.

LLMs are just one component of AI, and something that really doesn't change what is needed long term. This is a great buying opportunity for AI stocks.

 

Yeah no doubt this was intentional for multiple reasons. Great time to buy in the dip.

 

so who had the big short on nvidia?

 

if their model required significantly less chips, it is game changing. but we don't know how many chips they are runnign to train it as they aren't going to admit to it.

I have to think that nvidia somehow has a way to know appx how many chips the chinese are running. reports are exponentially higher than they are claiming

 

It's possible they had fewer chips because it's mathematics LLM. Without going into crazy machine learning details, math is about the easiest thing they can train because it's logic based. Even if it's true, I doubt that applies to the entire industry.

 

Yeah, like 90% of the coders and IT people at the mega company where my son works are Asians ... it's not hard to believe they are capable. They just try to come to America because we pay better, so maybe we are under a false impression that Americans are good at this stuff.

 

It's not based off of open source, it IS open source. You can download its source code and see for yourself if it's a clone of Llama. You can even read the paper they published on how its made, complete with all the weights and the formula they used, and build it yourself from scratch to see if it really does only need the number of chips they said they needed. This would still require many millions, as while the training run only cost them $5.6 million, the capex needed to buy the chips would still cost many more millions (reportedly ~2000 H800's at ~$20k each, so ~$40 mil).

With that said, you can bet your bottom dollar that many companies around the world will be doing exactly that. Many companies have stayed out of the LLM training business because the assumption is that it takes billions to buy the necessary chips and tens of millions if not hundreds of millions to do a training run. What Deepseek has done is to prove that you can spend just tens of millions to buy the chips and a few million to do the training.

Do you know any elderly folks who don't know how to use a smartphone and feels life is perfectly fine without them? Do you remember when the internet was just becoming a thing and people wondered why you would use that and not just read books and newspapers? AI can do the things you listed but already so much more.

For example, do you know how to make a website? If you're savvy with AI you can make one in hours whereas before it would take you at least weeks to learn. Do you know how to write a funny but compelling essay? It may take you years to never learn how to do so, but if you're savvy with AI you can write one in minutes. You can have AI make Excel spreadsheets, Powerpoints, make posters or Ads, plan a trip to Italy, or just a fun family photo for your annual Christmas card. You can be a real Renaissance man just by learning how to use AI.

Imagine having a conversation with a guy from 1985, with you having a smartphone with access to everything on the internet in 2025. Do you know how much smarter you would sound? It'd seem like you're dropping knowledge left and right, and fact checking him every step of the way with dead on accuracy. You'd sound like an encyclopedia to him. This is what being savvy with AI would be like right now if not in the near future. You'd seem like a scientist, engineer, and playwright to the unsavvy. The things I listed previously are just minor consumer uses, imagine these things being used for actual productivity, which it already is being employed to do so. Billions have been invested in AI, yet tech companies keep cutting jobs, don't you wonder why?

 

appears that a difference (maybe significant) between Deepseak and ChatGPT is a mixture of experts approach versus transformer approach.

 

Will never know but I’d assume it’s a mixture of hedge funds. Also NVDA has multiple leveraged ETFs so the unwinding of positions can really snowball. I’m sure there were also a ton of retail that was over concentrated/leveraged in the position; late as usual and panicked.

 

logging keystrokes and saving the data on chinese servers..all those US downloads..US regulators asleep at the switch allowing it to be doenloaded in the US with those privacy conditions in the T & C's. even an idiot like me could see that coming

but hey, we made google maps change the name to gulf of americuh...winnnnner

DeepSeek collects keystroke data and more, storing it in Chinese servers

U.S. tech companies are known to stockpile as much user data as they can, but DeepSeek's privacy policy makes Meta, Google, and OpenAI look tame.

Over the past few days, China-based AI startup DeepSeek has catapulted into tech consciousness with an open-source model that many claim is just as good, if not better, than OpenAI models and API costs for a fraction of the price. DeepSeek R1 might be significantly cheaper to run, but your privacy and security are the actual cost.

"The personal information we collect from you may be stored on a server located outside of the country where you live. We store the information we collect in secure servers located in the People's Republic of China," the privacy policy reads.

In another section about how DeepSeek shares user data, the company states that it may share user information to "comply with applicable law, legal process, or government requests."

 

Looks like the Chinese needed bullets for the AI gun and Nvidia was all too willing to help load it.


The Biden administration in 2022 put in place controls on chips exported to China. U.S. companies that wanted to sell to China first needed to throttle a chip function called interconnect bandwidth, which refers to the speed at which data is transferred.

In response, Nvidia, the world’s leading designer of AI chips, came up with a new product for China that complied with this parameter—but compensated for it by maintaining high performance in other ways. That resulted in a chip that some analysts said was almost as powerful as Nvidia’s best chip at the time.

U.S. officials vented publicly and privately that while Nvidia didn’t break the law, it broke the spirit of it. The government had hoped that industry leaders would be collaborative in designing effective export controls on fast-changing technology, said a former senior Biden administration official.

https://www.wsj.com/tech/ai/china-deepseek-ai-nvidia-openai-02bdbbce?mod=hp_lead_pos1

 

I wouldn't worry too much about that. APIs based on DS is popping up like wildfire, soon we'll have all sorts of non-Chinese APIs hosting data all over the world available for use. Already we have the Indian startup Perplexity running the R1 model for their search engine.

Right now it's the very early days for AI. The savvy can already use it for great benefit, but for general consumers it's more or less a toy. It's kind of like the early internet days, with the infrastructure growing exponentially but for the general consumers all they had was the Email. It'll lead to a stock market bust as we're seeing now due to the slow down in wild infra spending (GPUs hardware wise, LLMs software wise) and companies pivot to looking for the next big money maker, the next YouTube, Google, Facebook, etc.

A personal assistant app seems like a natural next step. A cross-platform app hosted on the cloud that can do everything for you. Schedule you doctor's appointments, book your travel accommodations, write your personals on dating apps, buy groceries (via delivery apps), perhaps even acting as your surrogate if you have an accident and is in a coma because it probably knows you better than you do. It could be rather dystopian to some, just like with smartphone and all the data they already gather on us, but this would be that on steroids. For ones who grow up with it though, life would seem so much harder without it.

 

Co-worker sent me this a lil while ago..

 

AI bot war?

 

this seems highly relevant.

AI can clone itself? SO China has model of chatgpt and asks it to clone itself and they are 90% of the way there?


Scientists warn AI has crossed ‘red line’ and can now clone itself

Scientists have warned that artificial intelligence (AI) has crossed a "red line," as the technology now has the ability to duplicate itself.

Researchers at Fudan University in China discovered that two large language models (LLMs) could self-clone.

In their experiments, the team used Meta and Alibaba’s LLMs in 10 different trials to determine whether AI could cheat. The results showed that the two models were capable of creating replicas of themselves in 50% and 90% of cases, respectively.

"Successful self-replication under no human assistance is the essential step for AI to outsmart [humans] and is an early signal for rogue AIs," the researchers wrote in their study, published on arXiv.

 

Not sure. He said it was probably hackers trying to steal the code, which I responded that it was like stealing from a British museum. I.e. trying to steal something that's already been stolen. Lol

 

and then there is this sweetener

another Open AI safety researcher leaves after 4 years in protest of shortcuts

good thign we ahve a crack legislative and regulatorys taff watching this stuff..

Latest OpenAI Researcher to Quit Says He's 'Pretty Terrified'

An OpenAI safety researcher announced that he left the artificial intelligence giant last year, noting that he is "pretty terrified by the pace of AI development these days."

In posts on X, formerly Twitter, Steven Adler shared that he quit OpenAI in mid-November due to the "risky gamble" the artificial general intelligence race is taking on.