It started with deepseek v3, which rendered the Llama 4 already behind in benchmarks. Adding insult to injury was the "unknown Chinese company with 5..5 million training budget"
Engineers are moving frantically to dissect deepsek and copy anything a...
One thing that I find worrying is actually that, although US IT companies are objectively and clearly falling behind Chinese ones, their stock price keeps going up, I suppose moreso due to their closing ranks with Trump.
Admittedly I know next to nothing about share value trading, but shouldn’t developments such as these temper investors interests in those specific companies making “AI” their next big thing? I’m afraid the bubble burst will be worse than 2008.
I very much expect the AI bubble to pop in the near future, and DeepSeek goes a long way towards ensuring this happens by completely breaking the business model of western AI companies. Nvidia and AI basically carried the whole market, and now it turns out that you can do the same thing with a fraction of the resources. That means nobody needs huge volumes of latest and greatest Nvidia chips anymore.
But what is to limit Nvidia from just shifting to “wow now that more can be done with less, we will do even more with EVEN MORE CARDS” and apply the improvements from Deepseek to their own models while continuing to ramp up in the same way?
Eventually there will be a pop, but I genuinely don’t get how this ushers in that pop. I guess the diminishing returns of these systems is hastened? As in, adding more cards with more resources will continuously have less effect. But companies will for now still want even the smallest edge by putting Deepseek-style-improved-AI on even bigger data centers.
That’s what OpenAI thought originally when they started working on ChatGPT5, they figured they’d just make the model bigger and it’s going to do more. Turns out that making the model bigger doesn’t actually produce better results. We’re also at a point now where most of the publicly available information has been scraped as well. Now the focus is turning towards improving algorithms for making sense of the data as opposed to just stuffing more data into the model. And this is a problem for Nvidia because current generation of chips is already good enough for doing this.
Of course, people will find ways to utilize more processing power as is always the case. But at least in the near term, this is no longer the bottleneck.
Interesting, I was unaware that we had already hit the limits of used data. That makes sense then. I hope that your analysis is correct, it’d be good news
the US gov can always cut off US companies from using the services of (much cheaper and better) Chinese AI companies. in fact I would be surprised if they didn’t at some point.
This is absolutely what will happen. I think it is advantageous that they don’t need to worry about any commercialization and also aren’t rushing to scale up like the US is, even more so now with project Stargate. I think there might be some benefits eventually from AI, but it makes sense to me to let it cook a little longer and be more methodical then powering up 3 mile island and filling it with GPUs
so all fiat is fictitious money they can pump it forever until a day comes where the US dollar is not desired because nothing useful can be purchased with it. the whole gov promoting crypto thing [insert Kamala twitter image of promoting crypto to black men here] is a way to strengthen the position of the dollar. money needs to be converted into USD in order to buy cryptos on exchanges. it’s continual increase in price makes capital want to enter it, which requires the purchase of dollars.
I’m not an economics guy or versed in the intricacies of stocks. So, that disclaimer out of the way, one of the things about the 2008 collapse is that there were assets that corresponded to the junk bonds etc that propped up the whole bubble. Houses were a big one, but also companies that made things like automobiles. Even the banks themselves served a purpose in the economy outside of their bubble-related shenanigans. Tangible assets and important roles. The necessity of these things was one of the justifications for the bailout.
AI doesn’t really make anything. There is no corresponding real world counterpart to an AI generated thing. It’s not like a car or a house or a plot of land or a potato. It exists in the wholly digital realm, suspended on threads of electricity. Oh sure there’s server farms, there’s electricity costs, there’s all the usual externalities you expect which are all real and tangible but there’s no tangible product at the end of the conveyer belt. There’s just more belt! It’s a machine that makes machines that make machines etc etc.
So yes we should all be very suspicious of this entire thing. While the techbros concentrate on how high they can stack the boxes marked “AI” we should be aware that the boxes are all empty. It’s actually kinda terrifying for 2 reasons. First, a giant financial collapse sucks to live through and I’d rather not do it again. Second, it is very scary the powers that be allowed all this to happen.
One thing that I find worrying is actually that, although US IT companies are objectively and clearly falling behind Chinese ones, their stock price keeps going up, I suppose moreso due to their closing ranks with Trump.
Admittedly I know next to nothing about share value trading, but shouldn’t developments such as these temper investors interests in those specific companies making “AI” their next big thing? I’m afraid the bubble burst will be worse than 2008.
I very much expect the AI bubble to pop in the near future, and DeepSeek goes a long way towards ensuring this happens by completely breaking the business model of western AI companies. Nvidia and AI basically carried the whole market, and now it turns out that you can do the same thing with a fraction of the resources. That means nobody needs huge volumes of latest and greatest Nvidia chips anymore.
But what is to limit Nvidia from just shifting to “wow now that more can be done with less, we will do even more with EVEN MORE CARDS” and apply the improvements from Deepseek to their own models while continuing to ramp up in the same way?
Eventually there will be a pop, but I genuinely don’t get how this ushers in that pop. I guess the diminishing returns of these systems is hastened? As in, adding more cards with more resources will continuously have less effect. But companies will for now still want even the smallest edge by putting Deepseek-style-improved-AI on even bigger data centers.
That’s what OpenAI thought originally when they started working on ChatGPT5, they figured they’d just make the model bigger and it’s going to do more. Turns out that making the model bigger doesn’t actually produce better results. We’re also at a point now where most of the publicly available information has been scraped as well. Now the focus is turning towards improving algorithms for making sense of the data as opposed to just stuffing more data into the model. And this is a problem for Nvidia because current generation of chips is already good enough for doing this.
Of course, people will find ways to utilize more processing power as is always the case. But at least in the near term, this is no longer the bottleneck.
Interesting, I was unaware that we had already hit the limits of used data. That makes sense then. I hope that your analysis is correct, it’d be good news
the US gov can always cut off US companies from using the services of (much cheaper and better) Chinese AI companies. in fact I would be surprised if they didn’t at some point.
This is absolutely what will happen. I think it is advantageous that they don’t need to worry about any commercialization and also aren’t rushing to scale up like the US is, even more so now with project Stargate. I think there might be some benefits eventually from AI, but it makes sense to me to let it cook a little longer and be more methodical then powering up 3 mile island and filling it with GPUs
Alibaba has been selling API access for their LLM for a while now. I don’t know how popular it is though. https://www.alibabacloud.com/en/solutions/generative-ai/build?_p_lc=1
so all fiat is fictitious money they can pump it forever until a day comes where the US dollar is not desired because nothing useful can be purchased with it. the whole gov promoting crypto thing [insert Kamala twitter image of promoting crypto to black men here] is a way to strengthen the position of the dollar. money needs to be converted into USD in order to buy cryptos on exchanges. it’s continual increase in price makes capital want to enter it, which requires the purchase of dollars.
I’m not an economics guy or versed in the intricacies of stocks. So, that disclaimer out of the way, one of the things about the 2008 collapse is that there were assets that corresponded to the junk bonds etc that propped up the whole bubble. Houses were a big one, but also companies that made things like automobiles. Even the banks themselves served a purpose in the economy outside of their bubble-related shenanigans. Tangible assets and important roles. The necessity of these things was one of the justifications for the bailout.
AI doesn’t really make anything. There is no corresponding real world counterpart to an AI generated thing. It’s not like a car or a house or a plot of land or a potato. It exists in the wholly digital realm, suspended on threads of electricity. Oh sure there’s server farms, there’s electricity costs, there’s all the usual externalities you expect which are all real and tangible but there’s no tangible product at the end of the conveyer belt. There’s just more belt! It’s a machine that makes machines that make machines etc etc.
So yes we should all be very suspicious of this entire thing. While the techbros concentrate on how high they can stack the boxes marked “AI” we should be aware that the boxes are all empty. It’s actually kinda terrifying for 2 reasons. First, a giant financial collapse sucks to live through and I’d rather not do it again. Second, it is very scary the powers that be allowed all this to happen.
They didn’t allow it to happen. It’s a fundamental law of capital. Even china has bubbles to some extent.
Well where else are they gonna put their money? Deepseek is open source lol.