After the layoffs, I became a tech godfather

Chapter 113: There is an even deeper abyss in the pit that is bottomless!

While Hao Cheng was exploring what changes Xiaosha's progress had brought, a large group of "Losers Alliance" and forces at home and abroad were making their move.

Xiaosha's update of the user agreement has caused a huge public opinion storm due to their constant speculation and deliberate interpretation.

As for the impact of this momentum on Xiaosha, at least for now, it has not had any impact at all, as the data shows.

There was no one who clicked to refuse and suspend the use of Xiaosha. Even for the two indicators of daily active users and average daily usage time of a single user, the data monitored by Baiju Technology showed no statistically significant changes.

Two whole days have passed, and Baiju Technology has not even made any explanation or clarification on this incident, which is enough to explain something.

Of course, the Losers Alliance and forces at home and abroad did not count on this - after all, let alone just interpretation, even if Xiaosha really has become an evil dragon, people still have to use it, there is no other choice at all.

What they need to do is to provide people with this choice at a time when the momentum is so great.

The timing was also very good. Public opinion had been fermenting for two or three days, and when it was about to decline, someone came to cheer it up.

A "thunderbolt" suddenly exploded!

At 12 a.m. on December 25 (24:18 on December 00, Pacific Time in America), when America should have been welcoming the birth of Jesus, OpenAI chose to work overtime and suddenly announced the release of a new model named: The New Humans.

They claim to have found a new AI training method that allows AI to learn knowledge and grow its abilities like humans.

The results are remarkable. After only one week of training, ChatNH1.0, the first-generation language model based on The New Humans model, has surpassed ChatGPT4.5 and even 5.0, which is still in pre-research.

OpenAI showed a capability growth curve of ChatNH1.0, with a parabola of time on the horizontal axis and capability on the vertical axis, which shot straight up into the sky.

Even people without any relevant professional knowledge can easily draw the conclusion that this model is getting stronger day by day, and each subsequent day's capabilities increase more than the previous day.

"Isn't this the English version of Xiaosha?" Many people think of Xiaosha when they see the name "new humans". Doesn't that look like a new human?
Add to that OpenAI’s introduction: “Learn knowledge and grow abilities like humans.” Isn’t that exactly the same as Xiaosha?
It has to be said that OpenAI has talented people. Just by naming it, it has made a strong connection with Xiaosha's achievements - as if all the abilities Xiaosha possesses are its own.

It would be fine if it were just that, it's just an English version of Xiaosha.

However, OpenAI then announced a major news: the new human model natively supports Chinese.

Moreover, they deployed servers in China in accordance with Chinese laws and regulations and trained the Chinese version of the new human model.

The New Humans is fully capable of providing continuous and uninterrupted services to all users worldwide. The Chinese version of The New Humans will be operated by Hangzhou New Humans Intelligent Information Co., Ltd.

"Oh no, this is aimed at Xiaosha!"

"What's wrong? Isn't it good for consumers to have multiple choices? China's laws and regulations do not discriminate. As long as the server is deployed domestically and the data is not transferred abroad, the big model can be operated domestically!"

"Okay, let's fight. I love to see rivers of blood!"

"Full competition is good for consumers."

On the Internet, a large number of supporters immediately emerged. This is not necessarily due to OpenAI's guidance. There are already many such people in China. Even voices criticizing Xiaosha have begun to emerge:

"Isn't Xiaosha awesome? With such a big public opinion and so many of us concerned, he didn't even respond. He really doesn't treat users as human beings! I didn't have a choice before, but now I choose ChatNH. Welcome, new human!"

"Xiaosha is no longer the only one. Even from an anti-monopoly perspective, we should support ChatNH. Competition leads to progress. We cannot rely on the kindness of others. We must take the initiative in our own hands. Only with rivals can it stop doing whatever it wants!"

"It's pointless to just argue. The game is now available for download in China. You can go and try it. The effect is really good, not inferior to Xiaosha at all, and it's completely free at this stage!"

……

Previously, public opinion attacks were indeed ineffective against Xiaosha because people could not do without it and had no other choice.

However, the attack from public opinion, coupled with the provision of alternatives, can be extremely powerful.

In particular, under the banner of being completely free like Xiaosha, it is inevitable that this first wave will attract a large amount of traffic. As for the result, it depends on how powerful these so-called "new humans" are.

"I wonder whether Xiaosha will live or die this time!" Hangzhou, Hangzhou New Human Intelligent Information Technology Co., Ltd. is also the qualification and operation company of the Chinese version of ChatNH New Human.

Originally, they registered a related company called ChatGPT Intelligence when big models were popular last year. Their original business was mainly to use ChatGPT's interface to re-develop some specific functions and then provide services to the outside world.

Just a few days ago, it was acquired by a foreign company, renamed New Human Intelligence, and hired a professional manager to be in charge of the company's work. This professional manager was none other than Wang Cheng from the previous band.

Wang Cheng is full of energy this time.

"But Mr. Wang, the capability of this model is only about one-third of Xiaosha's," the assistant reminded him.

"Does it matter? If you can't use it, the difference isn't that obvious. At least, it's much stronger than the previous ones like Wen Xiaoyan and Doubao. It's obvious that it's on the same table as Xiaosha. As for the lower ability, it doesn't matter at all!"

There is some truth in what Wang Cheng said.

In terms of capabilities alone, the New Human model is much stronger than several domestic models that also use the training method of [Reinforcement Learning Based on Xiaosha Feedback]. After all, OpenAI has a unique advantage, that is, it can stack computing power cards without restrictions, and buy as many as it wants.

However, these domestic companies do not have the conditions in this regard.

Although computing power cards are not completely unavailable, it is still a bit troublesome to purchase them on a large scale. Even if you can buy them, the cost will be much higher than OpenAI.

The so-called ChatNH, the new human, is the original algorithm. Combined with Xiaosha, it can provide unlimited feedback, and then accumulate computing power at all costs, which is as powerful as flying bricks.

It took two whole months to achieve this result. As for OpenAI's claim of one week, and the impressive curve, they are all lies.

Moreover, the ability diagram is portrayed as an absolute upward trend, which is intended to mislead people.

Everyone in the industry knows that there must be a limit, but they don’t know what it is.

As for Xiaosha, it is given. Even if you make all the sand on the earth into computing power cards and supply all the electricity to the computing power cards, and train them for a thousand years, you can only reach half of Xiaosha's level at most.

However, from the user's perspective, this gap is not as obvious as the data shows, especially when compared with the previous idiotic large model, it is definitely a rapid improvement.

"ChatNH is great, it's really useful beyond my expectation!"

"It's really useful. Let it talk to Xiaosha. If it can say ten sentences without revealing the truth, it's a talent! Then let it analyze what kind of job and position you are suitable for. The difference will be obvious at a glance!"

"But I just need a tool. ChatNH is free, so isn't that enough?"

"Exactly. And I always feel like being too strong like Xiaosha is not good either. Like a person, I always feel like someone is spying on me. I think ChatNH is just right!"

……

At this moment, users begin to disagree based on their portraits or positions.

Similarly, the data shows that the average daily usage time of users has decreased, and some users have even been lost.

"Mr. Hao, how should we respond?" Li Qingbo came over to ask for instructions.

"It's Christmas Eve in the US today, it's hard for them to still be working overtime at this time." Hao Cheng shook his head: "It would be bad if we don't express our gratitude, they will feel that they are not respected.

"First, block their accounts," Hao Cheng said. "Xiaoshan has the relevant account information. Ask Cao Song to restrict their login. Then release the relevant announcement to the public.

“Also, tell a story.”

"What story?" Li Qingbo asked puzzledly.

……

Hao Cheng paused: "Do you know what best represents love?"

"What?" Li Qingbo was stunned for a moment: "Rose?"

“It’s fried chicken!”

"What fried chicken?" Li Qingbo was still muttering to himself even after he walked out of the office. Especially after reading the short story provided by Hao Cheng, he was even more confused. What did it all have to do with that?

Li Qingbo didn't understand it because he was a businessman and didn't know much about technology, especially AI training.

As soon as Cao Song got the story, he understood it immediately: "Okay, this is not just a simple story. For AI, it is easiest to show its capabilities to the outside world. However, there is something more important than capabilities and levels."

"What is it?" Li Qingbo asked humbly.

"Rules." Cao Song said, "If a man is not disciplined, he will not be straight; if a tree is not disciplined, it will be weak. The ability of AI to be displayed externally is only part of it. The real core is that it must be controlled, and these things cannot be achieved by others.

“Traditional AI uses artificial rules to achieve control. For example, if the national flag cannot be removed, a rule will be established so that the flag-related capabilities will not be executed.

“For example, if a user says that he wants to commit suicide, the AI will certainly stop him. When it detects the keyword “suicide,” it will execute the predetermined code and logic instead of the AI’s own capabilities and logic.

"But AI like Xiaosha is not like this. I have tested it and it is truly integrated with the AI itself. It is natural and knows exactly what it can and cannot do. And it does not show this to others, so others cannot learn it. Do you understand?"

Li Qingbo nodded. He understood a little bit. What this meant was that the so-called "new humans" were trained with Xiaosha, but they only learned three-quarters of the abilities, and self-control could not be learned through training.

"Fried chicken is the best representation of love. If it can't be explained, it's just a joke at best. But an AI that is so powerful and has no ability to control it could potentially send itself to hell."

After saying that, Cao Song shook his head and began to observe a moment of silence for OpenAI and the so-called “new humans”.

If the fried chicken incident was a pit that was dug inadvertently, then everything that happened afterwards was the deeper, bottomless abyss beneath the pit.

(End of this chapter)