Chapter 389: Left-right Brain Chip
“I know everyone wants to know what to buy to get rich, what will surge.
Various affiliated companies, but actually the association is almost zero.
This is just like, my mobile phone number uses China Mobile, so can you say I have a deep cooperative relationship with China Mobile? It’s indeed quite deep, after all, there’s a transaction relationship every month, and China Mobile can know my call situation.
But can this be called Lin Ran concept stocks?
Probably not.”
Lin Ran joked.
The bullet screen was uniformly sending: “Ran Shen knows, buying China Mobile right away!”
“I think everyone’s thinking can be broader, deeper, thinking about the essence of the problem.
L5 will have the most direct impact; no matter which method it adopts for popularization as I mentioned, insurance companies’ auto insurance business will definitely be impacted, and automakers will definitely want to eat this cake.
This is continuous cash flow, there will definitely be very intense game, even if insurance companies can still get a share of the cake, the cake won’t be as big as in the past.”
After Lin Ran finished speaking, he paused for a moment, because he felt this point was enough; he originally only wanted to talk about stocks that would fall, not those that would rise:
“Silly Girl? Isn’t the name Silly Girl quite interesting?
But everyone can rest assured, when it officially launches, the autonomous driving system might be called this name, but the model definitely won’t be; I don’t think Huawei would agree to calling it this.
Why only cooperate with Huawei? Can’t stand companies doing patriotic marketing like this?
I need to say something about this: first, we haven’t always cooperated with Huawei; Deep Red has the most trade relations with Tencent, and Apollo Technology’s upstream and downstream supporting enterprises currently exceed one thousand.
Huawei is just a manufacturer with whom we cooperate more closely in specific fields.
We have never excluded cooperation with any company; including the LOGO on our lunar rover is also Xiaomi’s, and I personally went to the moon with Lei Zong.
Secondly, regarding patriotic marketing, I think it’s a paradox: why can’t we do patriotic marketing?
I always felt that the Chinese Internet smeared patriotic marketing, as if patriotic marketing is inherently wrong.
I don’t know if Huawei does this; even if they do, it’s not a reason for attack.
The ones who love patriotic marketing the most are America companies: Jeep, Coca-Cola, Ford, Nike—all do it; Budweiser Beer even directly named a beer America, I saw it when I was in America.
(Budweiser’s beer named America)
Japanese companies, German companies, Korean companies, etc., all do this; there’s nothing wrong with it, no problem at all.
Any Chinese Company doing patriotic marketing, I think it’s fine.
Even equating buying our product with patriotism, I don’t think it’s a big problem, because Huawei’s profit is shared with employees, employment stays domestic, production and manufacturing, research and development, marketing full process stays domestic; I don’t think that’s a big problem.
Even for any other Chinese Company saying this, I don’t think it’s a big problem, because it’s a Chinese Company; buying its product creates tax revenue for China, makes contribution to China’s operation.
From a consumer’s perspective, I think automakers should all do patriotic marketing in the future; they can publicly disclose at press conferences how much tax revenue they contributed, average salary for employees, how many upstream and downstream enterprises in the supply chain they drove; they can compare this.
Whoever has better data can claim to be more patriotic.
This is just my personal idea.
I think the funniest thing here is that the rice fields on one hand accuse patriotic marketing, on the other hand call for less regulation, small government; don’t you think this is very contradictory?
They want to play the market’s role, yet think patriotic marketing shouldn’t be done; shouldn’t the consistent approach of head and butt be to respect patriotic marketing and call for the market to solve this problem?
If the market doesn’t buy it, patriotic marketing naturally stops.”
Lin Ran unleashed a barrage:
“Am I a fanboy? Of course not; Huawei has many behaviors I can’t stand, like the mobile phones they sell in telecom and mobile service provider stores: high price low spec, then some middle-aged and elderly consumers who don’t know the market, out of supporting Huawei and simple patriotic sentiment, buy these high price low spec products; I think this is very bad.
People support you, I very very resent such behavior; here I publicly call for Huawei to stop this practice, including other mobile phone manufacturers; if they have similar behavior, they should correct it.
I just saw the bullet screen asking: Ran Shen, isn’t what you said contradictory? You said patriotic marketing is fine, then accuse Huawei’s behavior?
No contradiction; I’m not a rice field, I didn’t say to respect the market; what’s wrong with my statement? I’m not double standards.”
After Lin Ran finished, he waved his hand: “Okay, stop, next topic.
I saw that regarding my call for automakers to do patriotic marketing and conveniently announce tax revenue, employee treatment, and supply chain upstream and downstream, many people approve; I see many audience say they are willing to support such patriotic data-good enterprises, haha.
Actually, there are many self-media, and so much public data online; we can design a model to calculate a patriotic index; if automakers don’t promote it themselves, we can measure it.”
Hearing this, the most panicked are actually not the automakers listed overseas; like NIO, although listed on NYSE, it’s all losing money, losing foreign shareholders’ money, a bit like Luckin Coffee, using foreigners’ money to do car manufacturing in China, not profitable yet.
The most panicked are joint venture brands, especially Japanese joint venture brands; these joint venture brands have a lot of suppliers that are Japanese companies; the leaders of these joint venture brands are panicking when seeing this, they already foresee the tsunami coming.
Originally these years joint venture brands haven’t been selling well; this stick won’t directly kill them off?
“Ran Shen, why not make chips yourself, but cooperate with Huawei?
I’m human, not god; Ran Shen is just flattery from everyone; I can’t do everything myself, no one can.
No company can do everything itself; NASA’s in the 1960s sent humanity to the moon, they had over 4000 suppliers.
Taking semiconductor as example, Huawei has advanced process experience, AI computing card business, most importantly secrecy; they fully proved their secrecy ability in that Redmond visit in 2023.
After all, I think whether superconducting chip or Apollo Technology’s capabilities extended from frontier materials, they are all very valuable.
Recently the joint production line Huawei and we built being secretly photographed also proves this.
So at that time when selecting core partners, it looked like countless choices, but actually really only Huawei could be selected.
What I can tell everyone now is that we are in talks with these four domestic computing card companies: Cambricon, Moore Threads, Lingxi Technologies, and Biren Technology.
Huawei is now working on the left brain, and we need companies to work on the right brain.
Bullet screen is already flooding the screen.
“Okay, I’m not recommending stocks. It’s not confirmed yet which specific companies we’re cooperating with, and it might still be handed over to Huawei. I haven’t decided yet.
Let me explain the logic to everyone. Humanity has a distinction between left and right brain, with relative advantages in division of labor and extensive synergy.
The functions of the left brain are language processing, logic analysis, sequential processing, and fine motion control.
The functions of the right brain are spatial sense, holistic processing, emotions and music, image thinking, and body control.
The brain connects the left and right hemispheres through the corpus callosum to achieve information exchange. Most complex activities require cooperation between the left and right brain, rather than being completed by one side alone.
The common saying that the left brain is rational and the right brain is perceptual, the left brain is science and the right brain is art, is actually an oversimplification.
In reality, each hemisphere can participate in multiple functions, but certain functions are more prominent on one side.
Similarly, why is it a language large model? Because the current GPU is more like a left brain chip, with logic, sequential processing, and precise calculation.
But relying solely on the left brain, you cannot truly achieve AGI, not even close to human thinking.
Whether it’s Deep Red, OpenAI, or Alibaba and ByteDance’s large models in research and development, I believe everyone is gradually perceiving that its speed of progress is slowing down, and it’s about to hit the wall ahead soon.
We’re now mostly doing cost work, figuring out how to reduce costs, how to lower the cost for users to call large models. Its ceiling is very obvious, right there.
Lin Ran pointed at the ceiling of his residence.
“In my view, it’s missing the right brain, missing the right brain chip.
Humanity’s left and right brain exchange information through the corpus callosum. This asymmetric function plus high-speed synergy architecture is the result of millions of years of evolution.
The example is right here on site.
Forget it, I’m in high spirits today, so let me tell everyone about my demonstration process, which is also my latest research achievement in the artificial intelligence field, on why we need the left-right brain structure.
“???”
“Ran Shen, don’t say it!”
“I don’t want to hear it!”
“Don’t say it!”
Chinese netizens unanimously said no no no.
“Ran Shen, if you really have to say it, wait until I clear out the foreigners and rare ones from the live broadcast room!”
Lin Ran saw it and laughed loudly: “Everyone relax, this isn’t any secret. I’ve been around the world for so many years, I definitely know what I can say and what I can’t.
The paper will soon be publicly published, including our requirements for right brain chip design, which have already been sent to the four manufacturers I mentioned earlier. Everyone can relax.
Logic tasks’ computational complexity tends toward low parallelism and high sequential dependency; perception tasks’ computational complexity tends toward high parallelism and low sequential dependency.
This is all the core.
Once everyone understands this sentence, we can continue.
“Oh no, I just realized I don’t have a brain!”
“This is a bit too difficult for non-computer professionals.”
“Didn’t understand”
Lin Ran looked at the bullet screen and smiled wryly: “Simply put, perception and logic have different requirements.
Alright, I won’t do too much derivation. Those interested can wait for the subsequent paper.
I basically started from the computational characteristics of tasks and analyzed the necessity of left-right brain chips.
In summary, Deep Red’s future breakthrough requires a breakthrough at the hardware level. Now, relying solely on algorithms and data has reached a bottleneck.
The optimizations we’re doing now can only reduce costs; the experience is already very hard to improve further.
Then the new hardware needs to have both left and right brain capabilities, and on a large scale.
The demand for right brain chips has been sent to the manufacturers. We are happy to cooperate with every Chinese company, note that it’s every one.
As for which of these four companies can become our supplier, I don’t know. This really depends purely on ability, purely on hard strength.
You first have to stand out among these four, then compete with Huawei, at least be on the same level as Huawei, and we will tend toward you for the purpose of balance.
What I want to say is, whichever of these four can break out, everyone can invest a little.
But I don’t know if these four have listing.
I wonder if everyone realizes the key point after hearing this?
That is, if it’s left-right brain chip, then the number of chips in the matrix will increase. Originally, maybe 200 cards to build a calculation matrix, now with left and right brain combined, you’ll need 400 cards. The performance in the physical world will become a shackle binding you, latency will become a hindrance to calculation. Even with the fastest interconnect bus, it has an upper limit.
And at this time, is there a possibility that to achieve AGI, you must have superconductivity?
