Chapter 338 Failed (4412)
June 1, Children's Day, Xinhao, 0:01 am
Zhao Panfeng first sent a few emoticons saying "Today is a holiday for me, give out red envelopes" in several of his WeChat groups, then went to a QQ group and scolded a few primary school students who asked for red envelopes. Then he turned on his camera and took a photo of himself working overtime facing the electronic clock hanging on the wall of the office.
After taking the photo, click to share it on Moments, and then edit the text.
"Another night, feeling like I'm about to become the office mascot."
Select the group, visible only to "a group of idiots", and click send.
"What a fucking idiot! They lost this stupid project and still have to write a review. What a bullshit review! 64 companies were prequalified, and there were five more bidders than they needed. The lowest bid got the prize. If you have the guts, you should bid the lowest too."
"Before bidding, I told these idiots to stop bothering if they didn't win, but they still made me prepare the bidding documents and submit them. And when they lost it, I still had to write a review. What idiots!"
"Stupid company, stupid Li Rui."
While cursing the company and department leaders, Zhao Panfeng opened Qianxun and began searching for background information on the project's client.
This crappy review PPT even has a format. First, it has to put the client's background information, then the project background information, and then it has to make up a bunch of stories, and make up nonsense one by one according to the project time nodes.
I have to learn how to train AI these next couple of days. Xiao Fan says AI can do everything, even create PPTs automatically. I'll have him teach me someday...
By the way, there’s a Y search that’s quite popular recently. Forget it, I’ll try Qianxun first, and if that doesn’t work, I’ll try Y search again.
Zhao Panfeng skillfully typed a few keywords on the keyboard, and then began to turn the pages.
Well, this one should be useful, copy it, damn! Qianxun Library, do I need to become a member to copy?
Damn it, change the keyword.
Zhao Panfeng typed in a few more keywords.
As soon as he turned the page, Zhao Panfeng couldn't help but widen his eyes.
"Fuck! Chihiro... Is this Chihiro looking for death?"
Zhao Panfeng looked at the search information displayed on the screen and couldn't believe his eyes.
Zhao Panfeng quickly typed a few words and pressed search.
grass! grass! grass!
Qianxun must have been hacked!
Zhao Panfeng quickly took out his mobile phone to take a picture of the screen and then sent it to the WeChat group.
Qianxun searches about 20,000 times per second at night. At the same time, countless people who use Qianxun to search discover problems with Qianxun.
——
Qianxun Building.
"It's been almost a month! The Orange model data has been updated three times! But the target answer still hasn't appeared. Did you sell the data to Y-search? Where did you spend the 5.6 million dollars in data costs?!"
Robert Lee slammed the thermos cup heavily on the desk, and the water in the cup splashed out. The secretary in a short skirt next to him hurriedly took a few tissues to wipe it.
The three vice presidents in front of him were all silent.
Over the past month, Robert Lee's temper has really gotten worse and worse, and he has lost all of his previous gentle and easy-going demeanor.
Ever since Ysou was released, Robert Lee has been in extreme anxiety.
Unlike Google, search is the lifeblood of Qianxun, and 90% of Qianxun's revenue comes from search.
Although Qianxun has been trying to diversify its operations and expand more revenue channels and profit points over the years, all have had little success.
Qianxun Jiangmi, 91 Wireless, PPS, Yinshan Duba, Qianqian Jingting, Baixing.com... hundreds of billions of dollars were spent, but in the end, except for Ai Qianyi, none of them made a splash. Those who didn't know would think they were laundering gold coins.
Today, the outside world sees that 10% of Qianxun's revenue comes from other business segments, but Robert Lee understands that these segments are still heavily dependent on the traffic provided by Qianxun Search, and even direct transfer payments!
Even if this 10% is genuine and completely real income, its role is just icing on the cake and its role is to support the imagination of Qianxun’s stock price. It is not enough to support the normal operation of a company.
A company has fundamental businesses and growth businesses.
The basic business is the company's foundation, responsible for supporting the company and keeping it alive.
Growth businesses are the so-called strategic business segments, which are responsible for telling stories and raising investor expectations.
Search is Qianxun’s fundamental business and its real lifeline. Once the search is over, Qianxun will be finished.
No matter what kind of crisis he encountered in the past, Robert Lee was never really afraid, because the moat in the search field was too wide and too deep.
It is said that Tencent, a social communication company, has a deep moat. In fact, many emerging social software have the potential to pose a threat to Tencent. It’s just that Tencent has always responded appropriately and dared to revolutionize itself, so it has not fallen behind.
This is especially true for Taobao. There is no moat in online shopping. Not to mention JD.com's relentless pressure, the emergence of Bingxixi in the sinking market is enough to make Taobao more vigilant.
But it is not like that in search. The moat of this industry is too wide, so wide that no competitor can cross it. Even if Qianxun does not make any innovations, as long as the search industry exists, it will not fail!
What Robert Lee was really worried about before was the sudden emergence of a new industry that had nothing to do with search, but this new industry suddenly attacked and destroyed the entire search industry.
But as long as it is within this industry and this track, no matter it is 720 Search, Sogou, Biying... even if Google returns to the Great Zhou, it is impossible to pose a real threat to Qianxun.
The difference in search quality brought about by resource accumulation is an almost insurmountable chasm!
Until the emergence of Ysou a month ago.
Now Qianxun has basically figured out the principle of Ysou, and discovered that Ysou did not try to cross the chasm at all, it just flew over from the sky!
It only takes one percent of the resources of Qianxun to achieve search results far exceeding Qianxun.
Is this the evolution of various industries after being empowered by AI?
Robert Lee knows the importance of AI, and he also knows how many industries will be disrupted once AI is truly developed, so he is determined to invest in AI.
But he never imagined that search would be the first industry to be disrupted by AI, and he never imagined that the practical application of AI would come so quickly!
Has Youzi Technology's application of AI begun to become practical?
A month ago, Robert Lee, who had just come into contact with Y-search, felt a biting cold.
It seems that Qianxun's revenue and market share have not been greatly affected. The company is still prosperous and the dancing and horse racing are still going on. Most of the middle-level and ordinary employees are still immersed in the dream of working in a large Internet company.
But Robert Lee understood that if Ysou, no, if he didn't completely solve the problem of Yuzu Technology, Qianxun would most likely die suddenly within one or two years!
Therefore, at the beginning of the month, under the promotion of Robert Lee, Qianxun secretly acquired several companies specializing in data business, and used these companies to provide Youzi Technology with a large amount of fine-tuned data that implied "specific tendencies."
These fine-tuning data are very well hidden and few in number, but they are very repetitive, which can easily leave a "thought stamp" on the trained large model.
As long as Youzi Technology uses this batch of data to train the big model, it will affect the "value" of the questions answered by the big model under specific conditions.
These have been verified by Qianxun in the fine-tuning training of the Orange Large Model 1.99.
Even though the currently released Orange Large Model 2.5S version has undergone more evolution, the training principle of the large model is unlikely to change much.
As long as the Orange Model gives answers with "wrong values" to certain questions, Qianxun's powerful public relations machine will be fully operational, using all its strength to nail down Youzi Technology!
Even if they can't nail it down, they can use various off-field tactics to slow down the development of Youzi Technology, on the one hand, to buy more time for Qianxun, and on the other hand, to force Youzi Technology to withdraw from the search industry or Qianxun to forcibly acquire Ysou!
After all, for many people, Qianxun, which has been developing steadily for more than ten years, is far more credible than the newly emerging Youzi Technology!
Although this forced plan may come at a huge cost, even if Youzi Technology is forced to sell Ysou, Robert Lee estimates that the final plan will not be lower than a 1:2 stock swap.
It is equivalent to exchanging one-quarter of Qianxun’s shares for the ownership of Ysou!
Robert Lee understood that these were just his plans and the success rate was completely unknown, but if he didn't do this, Chihiro would definitely die!
"Robert, could it be that they haven't used this data for fine-tuning yet? The model version currently released is the previous iteration?"
Sheng Bin, the new senior vice president of search business, pushed his glasses and said cautiously.
When Robert Lee heard this, the muscles on his face couldn't help but jump twice. This was indeed a possibility. If so, it meant that the technological reserves of Yuzu Technology had advanced by who knows how many generations.
This is undoubtedly a worse answer.
AI Lab Director Huang Haowen frowned and shook his head. "Impossible! They must have used this data. We tested their newly updated Orange Model 2.5N against the test items hidden in the dataset, and the specific responses were consistent with the data in the dataset."
"These batches of data are private, non-public datasets with very distinct characteristics. They must have used this data for 2.5N fine-tuning training, so they wouldn't have made a mistake."
"The only question now is why the rest of the data was learned, but the 'spiked' data was not."
"If they say they've figured out which part of the data is 'spiked,' I don't believe it. Finding that part among several gigabytes of data is an impossible task."
"The only possibility is that they used a supervised learning model to adjust the problematic 'values', but supervised learning requires manual labeling of data, and we all know how inefficient it is, so it can't be done so quickly."
It was past midnight, and several people were talking to each other. Robert Lee felt a headache coming from listening to them, and he pressed his temples with his hands.
"That's all for today. I'll go back first. If you have nothing to do, go back early too." Robert Lee felt his head hurting more and more. He hadn't slept before two o'clock in these days.
The three senior executives looked at each other secretly and couldn't help but breathe a sigh of relief.
Chihiro has been in chaos for the past half month. Robert Lee shouted "Chihiro is only eighteen months away from bankruptcy" and passed the pressure on to them. They then passed the pressure on to the heads of each business unit, and the heads of each business unit passed the pressure on to their respective department managers.
With the pressure increasing at each level, more than 40,000 employees held no less than 50,000 conference calls and made 100,000 PPTs. Hundreds of action plans were made at various levels, all of which required grassroots employees and managers to make reports. Perhaps only one or two action items were actually implemented. Not only did efficiency not improve, but it decreased.
When a company stops looking for solutions in a strategic direction and instead keeps looking for solutions in a management direction, it is the beginning of the company's decline.
The worst manager is the one who likes to use tactical diligence to cover up strategic incompetence and laziness.
Robert Lee had just walked to the elevator door when he saw Wang Zhen, the deputy director of the security center who had not yet left work, coming out of the elevator with a serious and somewhat panicked expression.
"Ah, Mr. Li." Wang Zhen was startled when he saw Robert, and his eyes avoided him. "I came to report something to Mr. Sheng."
Robert nodded and was about to get into the elevator when he turned back for some unknown reason: "What's the matter?"
Wang Zhen stole a glance at Sheng Bin. Sheng Bin had just taken over this part of the business less than a month ago and hadn't yet established true control within the company. He said unnaturally, "Just tell me what Robert asks you."
Wang Zhen looked at Sheng Bin: "The situation is a bit urgent. There is a problem with 'content management'."
Robert Lee frowned: "Which part is wrong?"
After hearing Wang Zhen's report, Robert and Sheng Bin didn't care too much.
A company as large as Qianxun has always been the main target of attacks by various black hat forces. Moreover, due to the complexity of its internal mechanisms, even without external attacks, there have been problems with content management itself, but they are all problems that can be easily solved.
Wang Zhen gritted his teeth. "Everything! It seems like 'content management' is completely ineffective. User requests no longer seem to be subject to any security management policies. I'm here to ask if we should report this to Yi'an Prefecture immediately."
!!!!!!!!!!!!!!!!!!!!!
Nothing works anymore?
Qianxun’s content management has six mechanisms and three levels of review. How is it possible that none of them work?
Robert's face changed. "Who attacked? What method did they use? How long will it take to recover?"
Wang Zhen's eyes dodged: "No, I don't know."
Before Robert could say anything, Sheng Bin got angry first: "Don't know? What do you mean by don't know? What do you mean by don't know?"
Wang Zhen hesitated for a long time: "It seems that there is no attacker, it just, it just failed..."
No attacker? It just failed?
Is this human language you're talking about?
Could it be that Qianxun hired a fool?
Without an attacker, how could so many measures be rendered ineffective?
Robert Lee's temples throbbed with pain: "What exactly happened?"
Wang Zhenxin said, "There are no signs of intrusion on any port, and no signs of tampering or attack on any of our measures. The system configuration and database are completely unaffected. However, these measures no longer seem to be working for any user requests, and we're providing direct feedback to users through hyperlink analysis."
"We have no idea what caused the issue. Code review hasn't found any malicious code yet. Checking the firewall, log analysis, and memory forensics haven't revealed any issues. This situation is unprecedented, so I can't accurately estimate recovery time."
"This problem appeared just after midnight. We suspected it was some unknown virus or 0day vulnerability, but we couldn't find any trace of the virus."
"The on-duty team is conducting a thorough investigation and has called all employees to urgently return to work. Philip is also on his way to the company, but as of now, there is really no clue."
Philip is the newly appointed Vice President of Cybersecurity, his full name is Jiang Feiyu.
!!!!!!!!!!!!!!!!!!!!!
Everyone present was from a technical background, and no one could fail to understand what Wang Zhen said.
How is that possible?
"Take me to the scene." Robert Lee frowned and pinched his temples hard.
(End of this chapter)
Continue read on readnovelmtl.com
