Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Chinese chipmaker shares surge 470%

bbc.com

Post reply on HN
Print view View original

181–190 of 198 posts

Re: Chinese chipmaker shares surge 470%

#181

Post by hn_submit »

glimshe wrote: post #2

I have nothing to say about this company's long term potential, but I'm curious to know whether this will become an enormous "tik stok" like SK Hynix became in South Korea. I've watched a short documentary about the investment mania in SK and it's really scary.

The DRAM market is a notorious pork cyclic business. SK Hynix and Micron have stood on the brink several times.

No matter what happens the Chinese will carve out a niche since they're playing the long game. And DRAM will always be needed.

Re: Chinese chipmaker shares surge 470%

#182

Post by tacitusarc »

NaiveBayesian wrote: post #75
esterna wrote: post #42

Earlier quoted context omitted.

Once I start a conversation about Postgres query plans, maybe 95% of the knowledge will be almost certainly not needed, and for an inference provider there will be many more concurrent queries with reasonably large overlap. Maybe future architectures will be able to take advantage of this so not all parameters are needed in memory for almost all instances. If we are talking about all knowledge , then I agree that the…

Mixture of Experts is already used by pretty much all modern LLMs to address exactly this phenomenon. Hopefully, future models can be trained to be even more aware of external knowledge, accessible through web search / RAG / whatever it will be then, and might not need to internalize much knowledge at all.

Note that MoE is a sparsification mechanism and doesn’t actually have to do with expertise or what would commonly be considered areas of expertise.

Re: Chinese chipmaker shares surge 470%

#183

Post by throw293745 »

pmxi wrote: post #145
mrtksn wrote: post #113

Earlier quoted context omitted.

No I don't mean anything related to politics of China-Taiwan. I have no interest in that.

Clearly you do, given you expressly maintain that China includes Taiwan (an uncommon opinion)?

Taiwan's official name is Republic of China.

The other China's official name is People's Republic of China.

Re: Chinese chipmaker shares surge 470%

#184

Post by drdrek »

xyzsparetimexyz wrote: post #27
fooker wrote: post #9

Prediction - we are going to figure out SOTA AI performance without requiring 1TB of memory within a year or so. Of course CXMT, Micron, and family will still be profitable, but maybe not 'surge 470% from IPO' profitable.

How would we do that? There's no historic precedent for that. Its fundamentally an information theory thing: what's the max amount of intelligence you can get out of 1 KB/MB/GB? There has to be a limit and I'm not convinced that it's far off.

But why do you assume that you need unlimited amount of intelligence to perform work? If you can cut out just the parts that are required for doing SDR work or Development, etc you can get performance beyond what we have with generalized intelligence. Even on the physical aspect think about it, if you need to do less networking between GPUs and can do more in each single chip, you can use more "tokens" for reasoning without increasing time/cost.

Re: Chinese chipmaker shares surge 470%

#185

Post by Mainan_Tagonist »

shimman wrote: post #139
Mainan_Tagonist wrote: post #80

Earlier quoted context omitted.

The stock market works that way, it's true. But what about SpaceX, Blue Origin... there are also counter examples that show it is possible for investors to wait out the big profits.

Both those companies only can exist because of massive amounts of corporate welfare. What exactly are you trying to argue here? That the government should give private capital more of the people's money while the country suffers from lack of healthcare, childcare, education, and housing? What's the point of this? Honestly?

The point I'm trying to make is that there are also investors that are long on an investment and willing take losses for a long time before the big pay off materialises, or not. But that's private equity investment rather than stock market logic. Capitalism can also be about the mid-long term profits, given the proper opportunity.

Musk did not go it alone in the SpaceX venture, he obviously had other investors joining in on the risks and potential losses.

As for the government giving private capital public funds, no, that's obviously not what i'm arguing for, even though that's pretty much how the US has been running since WW2. The military industrial complex did not rise up as a normal capitalistic venture.

I am as appalled as anyone by the divestments in healthcare, education and such in the US, but the moral underpinnings and fundaments of the United States and its history, the puritanical element in particular, lead me to think that this is going to remain a long term issue in this country.

Re: Chinese chipmaker shares surge 470%

#186

Post by Mainan_Tagonist »

yuraoak wrote: post #129
Mainan_Tagonist wrote: post #80

Earlier quoted context omitted.

The stock market works that way, it's true. But what about SpaceX, Blue Origin... there are also counter examples that show it is possible for investors to wait out the big profits.

those are not run by Wall Street. they are run by 2 billionaires who care about something other than near-term profit (legacy, ambition, etc).

That's exactly my point, those are run by private equity, investors willing to take the risks and wait the big payday out.

Re: Chinese chipmaker shares surge 470%

#187

Post by Mainan_Tagonist »

fwip wrote: post #133
Mainan_Tagonist wrote: post #80

Earlier quoted context omitted.

The stock market works that way, it's true. But what about SpaceX, Blue Origin... there are also counter examples that show it is possible for investors to wait out the big profits.

I don't believe either of those two companies have ever been profitable. (I'm sure Musk has some accounting tricks to make it look that way, though.)

Indeed, if these are loss making operations, who puts the capital for those rockets to get off the ground? and why are they putting their capital in this when instead they could have a "nice ETF bringing 10% yearly"?

Re: Chinese chipmaker shares surge 470%

#188

Post by andrekandre »

lelanthran wrote: post #69
t-sauer wrote: post #40

Earlier quoted context omitted.

Your example also counters your point: A lot of people do not care about 4k, and even less do care about 8k or above. We reached a point where most people are completely happy with the quality they get.

> Your example also counters your point: A lot of people do not care about 4k, and even less do care about 8k or above. We reached a point where most people are completely happy with the quality they get. And when (not if) we get to the 4k (or 8k) equivalent of LLMs, they'll just be baked into hardware and we'll have near instant responses while running locally. There is no future where OpenAI, Anthropic, etc survive…

  > a point where training a new model is done only every 5 years or so
not to take away from your point, but 5 years is a long time to be out of date (esp with programming languages)... or maybe im misinterpreting?

Re: Chinese chipmaker shares surge 470%

#189

Post by mortarion »

jonkoops wrote: post #7

EU chip industry: best I can do is eat glue in the corner.

It's simply not possible to build profitable fabs, or large industry anymore in most of the EU. This won't change. Laws and regulation has put a stop to this. The best that EU can do is to protect existing industry.

Re: Chinese chipmaker shares surge 470%

#190

Post by xyzsparetimexyz »

NaiveBayesian wrote: post #75
esterna wrote: post #42

Earlier quoted context omitted.

Once I start a conversation about Postgres query plans, maybe 95% of the knowledge will be almost certainly not needed, and for an inference provider there will be many more concurrent queries with reasonably large overlap. Maybe future architectures will be able to take advantage of this so not all parameters are needed in memory for almost all instances. If we are talking about all knowledge , then I agree that the…

Mixture of Experts is already used by pretty much all modern LLMs to address exactly this phenomenon. Hopefully, future models can be trained to be even more aware of external knowledge, accessible through web search / RAG / whatever it will be then, and might not need to internalize much knowledge at all.

I suppose the right approach would be to run queries like the ones you want your model to be able to do, sort weights/experts by usage frequency and then reareange the weight so that all but the most frequently used can stay on disk. Tricky though, it could be that a e.g. database question uses 90% of the network at some point or another

Edit: I guess at the moment this is just having an LRU cache of experts

Post reply on HN