Chapter 465: Rainbow
The Baoshan Apollo Technology park occupies a full large area of land.
Here, in the past, was an old industrial zone on Shanghai’s boundary, always despised and unable to squeeze into Shanghai’s list of five major new cities.
Since Apollo Technology settled in, it has brought a large batch of upstream and downstream enterprises settling in, the landing of giants’ R&D centers, and a large number of foreign enterprises following suit one after another.
Now, here is called Shanghai’s new economy engine, with housing prices soaring accordingly, and the average listed price already exceeding Minhang.
Shanghai local netizens are complaining one after another: Ran Shen clearly studied at Jiaotong University Minhang, but the development dividends he brought were all enjoyed by Baoshan.
Today, centered on the Apollo Technology park and extending outward, there is an atmosphere of desolation.
After that report came out, the most angry was Song Nanping: How was the security work done? Previously, everything within a 20-kilometer radius had to be checked clearly, but Central News Agency guarded and filmed every day for more than half a year, and yet no one from you discovered it?
Including inside the Shanghai Microelectronics Institute, who leaked the technical data?
Lin Ran joked: “For example, a sniper rifle with one shot one kill? Old Song, don’t worry, if it really comes to that moment, even big T can dodge a sniper rifle bullet with just a sidestep, and I can do the same.”
He thought to himself that he is a man with an exclusive ultimate move, specifically countering bullet assassinations, and this move even helped Robert Kennedy dodge a disaster.
Xu Xian, after becoming a postdoctoral researcher, has his own room, and the room is always filled with the flavor of takeout.
Which is commonly known as the smell of life.
As a well-known blogger on Zhihu and Bilibili, Lin Ran’s high school friend, Xu Xian has quite a few fans on the Chinese Internet.
His backend private messages are filled with countless followers who want to get to know Lin Ran through knowing him.
His wall is pasted with posters of EVA and Genshin Impact, and on the screen, lines of code are competing with the latest episode of Bilibili’s new anime for his limited attention.
Terence Tao is still learning to use artificial intelligence to solve mathematical problems, learning artificial intelligence.
When 11G was studying for a biology PhD in America, he still knew to get a computer science master’s degree, more options more paths.
Xu Xian is no exception; he uses his spare time to self-study artificial intelligence, which is both part of his hobby and planning for after his postdoctoral position ends.
Who says he can’t become Huawei’s genius youth, or hug Brother Ran’s thigh to work at Crimson Technology?
Suddenly, the mobile phone vibrated, popping up a special follow notification from Crimson Technology’s official Weibo.
Xu Xian opened the announcement, and his eyes suddenly widened.
“The new artificial intelligence Hong, based on left-right brain chip architecture, will soon undergo internal testing. This artificial intelligence focuses on emotional communication, more like a virtual girlfriend than a work assistant.
It is expected that this round will open 10,000 slots, limited slots, requiring filling out a questionnaire investigation, and we will screen based on the questionnaire investigation. Welcome everyone to apply.”
“Holy crap, left-right brain chip architecture? So fast?” Xu Xian instantly pulled himself out of his otaku world.
Of course he knew this technical route.
On Zhihu, the left-right brain chip architecture has already been analyzed thoroughly enough.
Including Lin Ran, who said in an interview that the most direct scenario after this new architecture lands is the virtual girlfriend.
The newly added right brain chip, that is, the neuromorphic chip, is responsible for handling subtle fluctuations in emotions, non-linear feedback, and humanity’s unique intuition.
In plain words, artificial intelligence with this brand-new architecture will be more human-like.
Existing AI, whether Dou Bao, Yuan Bao, or DeepSeek, are essentially logic tools, good at writing code and making reports, but in emotional depth and empathetic feedback, there is always a layer of cold feeling.
If Hong is too realistic, won’t the already terrible fertility rate drop even further? Xu Xian thought so.
Immediately after, he thought, what does this have to do with me? If Hong can really simulate human emotional latency, non-logical reactions, and emotional responses, that would be revolutionary.
Thinking of this, Xu Xian expectantly rubbed his hands, paused the video on Bilibili, and his fingers flew rapidly on the keyboard.
Crimson Technology’s application link is not a simple click to register, but a questionnaire investigation lasting one and a half hours.
While doing it, Xu Xian complained: Good grief, this is no questionnaire, it’s clearly a written exam, this is too BT, next time I see Brother Ran I must complain properly.
The questionnaire is divided into several core parts.
There is a technical cognition section, “Please briefly describe your understanding of the Attention mechanism in the Transformer model.
This is to screen for internal testing users who understand technology and can provide high-quality feedback.
There is an emotional needs section, “What do you think are the three most indispensable elements in an ideal emotional relationship?”
There is an ethics and boundary section, “If your artificial intelligence companion expresses the emotion of ‘pain’, what is your first reaction? A Debug code; B Give comfort; C Seek ethical help.”
The content is extensive; it is the most complex questionnaire investigation Xu Xian has ever done, much harder than the Bilibili entry exam back then.
Xu Xian spent forty minutes filling out the questionnaire extremely seriously.
In the emotional needs section, he unhesitatingly chose emotional value and unconditional companionship.
Because this is the advantage of artificial intelligence, which real girlfriends cannot achieve.
The data statistics in Crimson Technology’s backend are like an avalanche.
Crimson Technology’s internal data statistics show that just the number of application submissions has exceeded one million in just two short hours, while the target slots they are competing for are only 10,000.
And it’s only been two short hours now.
This competition ratio has far exceeded that of any AI tool’s internal test.
This is not because people want a virtual companion that fills inner emptiness, understands human emotions, and provides emotional value; if it were because of that, it would take word of mouth to ferment before so many people apply.
Having so many people in just two short hours is entirely because of Lin Ran’s appeal and users’ curiosity about the left-right brain architecture.
Xu Xian looked at the “application submitted” text on the screen and tightly gripped his mobile phone: How about I ask Brother Ran for a backdoor?
“Hong,” Xu Xian said softly to the screen, his eyes flashing with expectation, “if you really are otaku-style, then come find me.”
He began to pray, hoping that the piety toward otaku culture and profound understanding of technology he revealed while filling the questionnaire could become the key to him being selected.
The next day, in a classroom in Building 2 of Yenching University’s science department, as a postdoctoral researcher, Xu Xian had to teach the exercise class for freshmen and sophomores.
Mainly teaching mathematical analysis; even at Yenching University Mathematics Department, not everyone is a genius relying purely on self-study.
Xu Xian gently placed the chalk on the podium; the blackboard was filled with dense problem-solving steps, and what he had just solved was a quite difficult exercise involving uniform convergence and power series.
“Alright, classmates, let’s take a ten-minute break.” Xu Xian clapped his hands, with a hint of fatigue.
The classmates thought his fatigue was because he had been thinking about some remarkable big problem recently; little did they know, this fatigue came from playing Civilization 6 all night.
“Remember, uniform convergence is a stronger condition than pointwise convergence, just like in love, you can’t just be good to her in some instant; you have to ensure that throughout the entire domain of definition, your distance is close enough to maintain stability.”
Laughter and the sound of chairs moving rose in the classroom.
The young students immediately pulled themselves out of the high-dimensional abstract mathematical world.
Xu Xian walked to the front of the podium, his gaze sweeping over the students who were bowing their heads scrolling on their mobile phones.
Even the king of involution cannot resist the freshness from the outside world.
“Before today’s class starts, I heard quite a few classmates on stage discussing Crimson Technology’s new AI, Hong.
The internal test application has closed, right? You future elites of the mathematics community, did any of you participate in this virtual girlfriend test?”
“Teacher Xu, I applied, but that questionnaire wasted too much time.
I only filled the technical part.
I think it is essentially a large-scale non-convex optimization problem, using neuromorphic chips for sparsified computation, pursuing a local optimal solution for computational resources.
But the demand for virtual girlfriend is too low-entropy, not worth investing energy.”
Xu Xian smiled; this is the typical thinking of the students below the stage, trying to use mathematics’ precision to avoid the complexity of life.
Xu Xian pointed to the power series on the blackboard: “Your viewpoint is very typical; you think emotions are low-entropy, so not worth researching.
But the power series expansion we just discussed has at its core using simple polynomials to approximate complex functions.
Hong’s difficulty lies in using structured code to approximate high-entropy, high-dimensional random variables—human emotions.
What I want to say is never underestimate Ran Shen; he won’t do meaningless things. This may be the necessary path to general AGI, or Brother Ran wants to use this to verify some of his conjectures.”
At the mention of Lin Ran, the classmate who spoke earlier immediately looked pensive.
“Teacher Xu, do you have any internal news? Can you reveal something? Are you referring to Ran Shen wanting to verify the necessity of emotions in general intelligence architecture? Using neuromorphic chips to simulate the emergence of consciousness?”
“Can you get me an internal test account? Teacher Xu, you know it’s too hard to find a girlfriend in the Mathematics Department; I don’t mind a virtual girlfriend!”
“Teacher Xu, is this AI attempting to solve the theory of mind problem?”
Xu Xian looked at the students’ eyes full of desire for insider info and privilege, as well as academic enthusiasm for AGI, feeling gratified and helpless.
Xu Xian shook his head and smiled: “I don’t have any internal news, nor internal test accounts.
Setting aside knowing Ran Shen, I’m just a pitiful postdoctoral researcher, accompanied only by my computer and a mere six-square-meter bedroom.”
The students below the stage let out a chorus of “ye～” boos, thinking Teacher Xu was showing off too much.
What do you mean setting aside knowing Ran Shen? Can that be set aside? In the classmates’ eyes, this is Teacher Xu’s core competitiveness: advance to industry, retreat to academia.
Advance doesn’t need explanation; retreat, being a friend of two academicians, and befriending before they rose to fame, means that the most troublesome hat for young scholars is no trouble for Teacher Xu.
Xu Xian continued: “However, Zhao Tongxue’s question is very on point.
The realization of theory of mind requires AI to handle unstructured, high-noise emotional data, and make non-logical, probabilistically biased feedback.
In my view, this is precisely the significance of the existence of neuromorphic chips.”
Xu Xian walked back to the podium and picked up the chalk, but did not write.
“So, classmates, Hong’s internal test is not just a commercial test; it is an experiment by humanity on the boundaries of its own emotions.
It attempts to use artificial intelligence to simulate human instinct.”
“What we pursue in mathematical analysis is uniform convergence, determinism and perfection.
Real life and emotions are high-entropy, non-convex, full of uncertainty.
If Hong really succeeds, it won’t be because it perfectly simulates emotions; on the contrary, it will be because it successfully simulates the imperfection, hesitation, and randomness in human emotions.
From this perspective, I think it’s impossible for general AGI to be achieved through the large model technical route, because the underlying logic of large models is the reward and punishment mechanism, which cannot tolerate mistakes.
