Chapter 346: Deep Red And Wenxin Yiyan
“Not worse than GPT?” Xu Xian was a bit surprised after hearing it, but then thought that since it was Brother Ran’s masterpiece, he wasn’t surprised anymore.
Being able to land on the moon in one year, leading China’s version of GPT, is that strange?
Novels need logic, but reality never needs logic.
Top-tier geniuses in reality need logic even less.
“Like this, send me your resume. I can only say one sentence to Brother Ran for you: This is my university roommate, give him an opportunity at the same level.
I can’t guarantee more than that.”
Xu Xian said after a moment of thought.
Zhao Songxia said in surprise: “Thanks, Xu Xian. Next time you come to Yanjing, I wouldn’t mind playing the White Snake Lady.”
He knew very well that this was already the limit of what Xu Xian could do. If the other party said something like “leave it to me,” he wouldn’t dare to believe it.
Xu Xian knew the other was joking and said speechlessly: “Get lost, get lost. It’s a small matter. I’m not sure if Brother Ran will see it, or if it can help you.
Once you’re sure you can stay with Brother Ran, prepare the props then, okay.”
Zhao Songxia said: “Xu Xian, you’re the real pervert. Props are here!”
Xu Xian explained: “I meant cosplay props. What were you thinking!”
Later, every initial employee who stayed at Crimson Technology was poached by Zuckerberg with a starting annual salary of tens of millions of US dollars plus a green card. It was nicknamed the cradle of billionaires.
Of course, no one knows now that employees of future large models will be so valuable.
And this future is not far away.
Later, Xu Xian realized that his few words seemed even more valuable than a thousand gold pieces.
Having gotten Xu Xian’s promise, Zhao Songxia began preparing his resume. He planned to attach a review on artificial intelligence that he wrote himself, as a demonstration of his abilities.
Just a resume alone was too thin. Even if it could increase the success rate by half a point, he didn’t mind doing it.
For smart people, they know when to go all out. It’s fine to slack off usually, but once an opportunity arises, they absolutely cannot miss it because they didn’t try their best.
Meanwhile, Xu Xian was waiting for Zhao Songxia to send him the resume before forwarding it to Lin Ran.
While browsing Zhihu, Xu Xian once suspected that Zhihu was eavesdropping on his call. He had just finished chatting with Zhao Songxia when Zhihu pushed “How to evaluate GPT, why wasn’t GPT born in China” to his homepage.
This was an answer with over ten thousand likes:
“Thanks for the invite. I can only say that in leading technology, America is still deservedly the world’s number one, and in leading technological development, America is still the lighthouse.
You can say America is showing fatigue, empire’s twilight glow, its control over Earth far less than before, its geopolitical operation level also greatly diminished, Wall Street and Washington’s deep state government devouring this country, but you cannot deny that this country is still leading the world forward in technology.
Domestic early internet giants were all localized products of American companies. After the iPhone brought the smartphone era, domestic Android brands began to rise. After Tesla appeared, various domestic new forces emerged endlessly.
America pioneers, China follows, chases, or even surpasses. Currently, we extremely lack pioneering achievements from 0 to 1.
Take Ran Shen’s Apollo Moon Landing and Burning No. 1 as examples. Indeed, we achieved great feats that America couldn’t. We even stepped on the Lunar South Pole soil for the first time, leaving a permanent base there. I also don’t doubt that in the future, under Ran Shen’s cooperation with China Aerospace, it can be developed into a Moon Base.
But the former relied on America’s Apollo Moon Landing in the 1960s, the latter is a product after SpaceX’s Falcon 9. We have always been following.
Including in this aspect, America had GPS first, then we smoothly developed Beidou.
I also believe that in the near future, China’s domestic large model will be as good as GPT, or even better than GPT, but we are still following.
I’m not saying following is bad. Following is good, of course. Following others’ mature paths has the highest input-output ratio for this research method, very suitable for developing countries like us.
In other words, other countries want to follow China’s path but can’t.
I’m just always very curious about one thing: when America also runs out of tricks and has no new breakthrough technologies, what do we do? Can we really have endless originals?
I don’t like GZ or external network HGDs who always elevate the issue to a very high level. Because we don’t have it, Japan and Europe don’t either. Does that mean only America’s election is an election, and other countries’ aren’t?
I think the reason the most important technological innovations were first born in America is fundamentally two points. First, it has the world’s largest consumer market, enough population, which determines that your technological innovation has users to pay for it. In the 90s, Palm computers could sell millions of units easily. If placed in China at the time, a single unit at 5000 US dollars, let alone millions, selling ten thousand would be impressive. PDA laid the foundation for the smartphone era to emerge first.
The second point is that America has the world’s best capital market. After the Bretton Woods Agreement was overturned, excess US dollars continuously flowed into Silicon Valley, forming an unprecedented positive cycle: innovation-financing-listing-financial freedom. Even during China’s past nationwide entrepreneurship period, the capital market didn’t need innovation but making money. What is your business’s counterpart in Silicon Valley, how to localize it, why can you beat the giants?
The massive consumer market and the best capital market together created America’s innovation miracle. What we need to ask now is, how do we create our own innovation miracle?
Can we create the environment that America can create?
If not, then we can only rely on flashes of genius from geniuses like Ran Shen. I believe he will eventually bring unprecedented innovation one day, but can we keep producing geniuses of Ran Shen’s level?”
Of course, there were also some self-reflective answers from this country, which made Xu Xian angry just looking at them.
“Thanks for the invite. I’ve been researching artificial intelligence with Brother Ran recently. I just want to say four words: Just wait for Dragon’s Roar.
Just wait for Dragon’s Roar, okay!”
Because of his relationship with Lin Ran, Xu Xian was also a Zhihu celebrity, with fans growing from tens of thousands to hundreds of thousands.
Zhihu fans don’t inflate; hundreds of thousands already counts as a top blogger.
After his answer was posted, likes quickly broke a thousand. The comments in the answer were all:
“When is it coming? I can’t wait!”
“I almost forgot Ran Shen also studied artificial intelligence, PhD in Graph AI. Looking forward to Ran Shen giving us a surprise!”
“Dragon’s Roar again? Did I come late? Already prepared for the roar bang.”
“Ran Shen’s unofficial spokesperson, huh?”
But indeed, Dragon’s Roar came soon, just the first roar was from Baidu.
In mid-March, Baidu held a product press conference at its headquarters building in Yanjing. The spacious auditorium was packed, with hundreds of media reporters, investors, and tech enthusiasts seated below, expecting the surprise Baidu would bring.
On the stage, a huge LED screen flashed with blue and white lights, with “Wenxin Yiyan: Intelligent New Era” in the center of the screen.
Below the stage, front-row reporters were busy adjusting cameras, while back-row audience members whispered about the ChatGPT craze.
“Baidu caught up this time. Being the first in China to release an artificial intelligence large model, it wasn’t a waste to all in on AI for so many years.”
Baidu’s all in on AI was in early 2017, when former Microsoft executive Lu Qi joined Baidu as Chief Technology Officer. Less than 24 hours into the job, Robin sat with Lu Qi in front of the media for questions.
The strategy to all in on AI was then confirmed and first announced at the Baidu AI Developer Conference in July that year.
“Did Baidu collaborate with General Manager Lin of Apollo Technology? I saw online that General Manager Lin seems to be working on a large model too.”
“No, never heard of it before. But in the AI field, Baidu’s technical reserves, talent reserves, and deep accumulation, General Manager Lin probably can’t compare, since he’s just one person.”
“The most important is to buy Baidu stocks. Nasdaq opens tonight, Baidu stock will definitely surge.”
Baidu founder Robin, wearing a white shirt, stood in front of the Wenxin Yiyan press conference backdrop, with his signature smile on his face.
(Wenxin Yiyan press conference on March 16, 2023)
Robin looked passionate. In his view, this was Baidu’s comeback battle!
For months, the team worked day and night. Based on the ERNIE large model, with massive parameters, integrating knowledge graphs and multimodal capabilities, able to handle multiple complex tasks.
The internal executive in charge of development patted his chest and guaranteed to him: In Chinese capabilities, our Wenxin Yiyan absolutely doesn’t lag behind GPT!
But facing the pressure of the global AI competition, he knew this press conference couldn’t have any mishaps.
If the demo goes smoothly, it will be a milestone for Baidu AI; if there’s a glitch, the stock price will crash.
He took a deep breath and adjusted the microphone.
“Ladies and gentlemen, welcome to the Wenxin Yiyan press conference!” Li Yanhong’s voice was loud and confident, echoing in the auditorium.
Enthusiastic applause erupted below the stage, and the bullet screen was uniformly “China go, Baidu go” cheering.
He paused, then continued: “ChatGPT opened the generative AI era, but we at Baidu released the ERNIE model as early as 2019.
Today, Wenxin Yiyan is here! It’s not a simple replication, but an innovation based on the Wenxin large model, excelling in literature, logic, and multimodal generation.
Of course, it’s not perfect yet, but we will iterate continuously.”
Below the stage, Chief Technology Officer Wang Haifeng nodded, holding a remote control, ready to assist with the demo.
Lu Qi left Baidu in 2019, widely seen as Baidu’s self-rescue failure.
Wang Haifeng thought to himself: Our model leads in Chinese understanding because it incorporates massive domestic data, but the demo must be intuitive, letting ordinary people feel the advantage.
Li Yanhong walked to the demo table in the center of the stage, where a large-screen computer was placed. He tapped the keyboard, and a clean chat interface popped up on the screen.
“To save time, we prepared a pre-recorded demo video, but the effect is absolutely real.” He explained, then pressed play.
Soon the entire process greatly disappointed the audience below.
It was just showing a scripted video interacting with artificial intelligence.
And some questions raised in the video seemed very simple, solvable merely by typical search engine services, like: “Which region of China is the author of ‘The Three-Body Problem’ from?”
“Is Robin treating us like fools?”
“No, if it’s just these functions, why use AI? What’s the advantage of this AI over a search engine?”
“Baidu is getting unscrupulous to claim the title of the first domestic large model.”
This performance extremely disappointed investors. 10 minutes after the press conference started, Baidu’s trading on the Hong Kong Stock Exchange quickly dropped 10%, evaporating over 40 billion US dollars in market capitalization.
“This looks pre-recorded. So far it’s just PPT.
These demos don’t look spontaneous.
It neither cited anything extracted from their database, nor from their model, nor compared it to ChatGPT.
Everything is very vague and theoretical.
If two words to describe it, undoubtedly disappointment.” The reporter noted on the notebook computer.
On stage, Robin could feel the disappointed mood below, but he didn’t mind and still followed the established program to complete the entire press conference.
He bowed: “Wenxin Yiyan is now open for internal testing. Welcome everyone to apply.
In the future, it will integrate into the Baidu ecosystem and change our lives.”
Below the stage, under the lead of Baidu employees, thunderous applause erupted.
Robin breathed a sigh of relief inwardly: Now we’ve taken this step. We are China’s first!
“I can’t say we’re fully ready to compare Wenxin Yiyan with ChatGPT, or even GPT-4. That’s a very high standard after all.”
“No large tech company globally has done it. We are the first.”
“Because of market demand, Baidu’s various product lines—search, autonomous driving, Xiao Du, etc.—are all waiting to use such technology.
In the external market, clients and partners are waiting too, everyone is urging.
So we had to release it first at this time.
It’s not perfect, but it’s the first, the first domestic large model. We can soon provide it to partners and consumers. That’s the most important!”
Robin added during the reporter interview.
After everything ended, Robin called the Chief Financial Officer: “How is it? Have we found the analysts from America?”
“Found them. We separately engaged authoritative analysts from Citibank and American banks to test our Wenxin Yiyan. They will give positive reviews, and tomorrow the stock price will return to its proper position.” The Chief Financial Officer said firmly.
According to Baidu’s script, this is called starting low and ending high.
However, Tencent had no intention of letting Baidu have its way. They had decided to use Crimson to step on Baidu’s Wenxin Yiyan to rise, making Wenxin Yiyan a sacrifice.
“Ready?”
Crimson had no press conference, only a quiet launch.
