After the layoffs, I became a tech godfather

Chapter 156 Seems like science fiction, but is actually reality

After communicating with Hao Cheng, Yu Dong thoroughly understood Hao Cheng's ideas and his future goals.

"We cannot regard Baiju Technology as a company at all," Yu Dong said while discussing with He Gang after leaving the Innovation and Entrepreneurship Park.

"Using corporate logic to analyze the behavior and decision-making of Baiju Technology is absolutely the wrong approach."

As he spoke, Yu Dong turned around and glanced at the office building of Baiju Technology. He suddenly had a different feeling in his heart: With this company, wouldn’t the progress of the entire world be accelerated?
Once this idea came to my mind, I couldn't stop it. The more I thought about it, the more possible it seemed.

……

Meanwhile, Hao Cheng and Wei Shujie were also having a brainstorming session, discussing their ideas for the next generation of computing cores.

"I am sure that non-deterministic input can indeed make tensor operations faster and bring more possibilities to evolutionary reasoning."

During the weeks since meeting Hao Cheng, Wei Shujie has not been idle for a moment. He not only reviewed the relevant knowledge of fuzzy mathematics, but also consulted countless materials and carried out a series of hypothetical calculations.

This was the conclusion he finally came to.

Hao Cheng took a quick look and asked a question:
"The conclusion is indeed correct, but I have a question. There are steps that can quickly draw a conclusion using Xiaosha. Why did you choose to spend several days on your own?"

"I..." Wei Shujie hesitated. He didn't know how to explain this question.

For many problems, if you use Xiaosha's assistance, you can indeed save a lot of work. The results of several weeks can even be condensed into one day.

However, if this is the case, wouldn't it seem like I haven't done any work?
But you can't say that. You can't tell Hao Cheng: Are you doing this just to make yourself appear busier?
"Compared with AI, humans are indeed much weaker in capabilities." Wei Shujie organized his words and then continued:

“I have also read Baiju Technology’s articles countless times about the advantages of humans over AI and how to use AI as a tool, but in summary, the work that humans can do is too limited.

"This is still the case for highly educated people. If the scope is expanded to the entire population, this phenomenon will be more obvious.

"We all know that innovation, raising the bar a little bit, and extending a little bit more are simple principles, but how many people can really do this? Only one in a hundred!"

Wei Shujie mentioned a phenomenon that Hao Cheng had also considered before, but for which there is currently no good solution.

Many occupations today, such as assembly line workers, deliverymen, and taxi drivers, which are the basis of their survival, will be easily replaced by AI and robots in the near future.

The current solution is a passive response: do not develop these machines that replace a large amount of labor.

But obviously, this is not in line with the general laws of economic and social development - advanced productivity will definitely replace backward productivity, it's just a matter of sooner or later, and it may not even be able to be stopped by human power.

As for what Baiju Technology has always advocated: starting from the weak points of AI and making more innovations and improvements, as Wei Shujie said, this is not an easy thing and not something that most people can do.

"Even our education over the past few decades has not cultivated students' abilities in this area," Wei Shujie added, "We are more focused on training industrial workers.

“Not to mention undergraduates, even at the master’s and doctoral levels, groundbreaking talent and projects are extremely scarce.”

"This issue is too big to discuss." Hao Cheng shook his head: "But I have a groundbreaking idea. I'll just say it and you can just listen to it."

"Ok?"

Wei Shujie said so many things in the beginning, actually he was trying to make up for himself, but when Hao Cheng heard it, he frowned and thought for a while, and then he came up with a bold idea.

"We now say that AI is a tool." Hao Cheng continued, "But what if we don't regard it as a tool, but as an integral part of human beings?
"The problem now is that the interaction capabilities of the borderless system are still not strong enough!"

Wei Shujie didn't understand. How was this related to the borderless system?

However, as Hao Cheng continued to explain, his mouth opened and closed, and it never closed at all for the next half hour.

Hao Cheng's idea is this:
The current system, whether it is a traditional keyboard, mouse, touch screen, or a borderless system's hardware fusion perception of operation intention recognition, requires people to take action if they want to use it.

For example, if I want to search for some information, whether I go to the search engine or call up Xiaosha, it requires an interactive step.

"What if this step is omitted?" Hao Cheng fantasized boldly.

"Omit?" Wei Shujie was a little confused: "How?"

"Just like you think, retrieve relevant information from your memory," Hao Cheng said. "Let the world's knowledge and data be mapped into a part of your brain's memory."

"Ah?" Wei Shujie understood what Hao Cheng meant, but how could it be possible to obtain massive amounts of data and information like he obtained his own memories!

"It is indeed impossible!" Hao Cheng said, "Unless we can identify human consciousness and convert the data into consciousness that humans can directly receive."

It sounds like this is not a problem that can be solved in a short period of time, or even within dozens or hundreds of years.

"But," Hao Cheng continued, "this thinking and direction is a living target. Although we can't do it, it doesn't prevent us from taking a step forward in this direction."

Hao Cheng now has some ideas in his mind.

Imitate the process of human brain thinking, drawing memory and inspiration:

The brain thinks one thing, it goes like this:

我要做一道数学题（1+1=？）（1）→大脑“调用”相关的数学知识解题（=2）（2）

And now, the operation steps of the unbounded system are:
我要做一道数学题（1）→唤出小沙（2）→通过融合感知的操作或者传统的键鼠语音等操作将这道题描述小沙（3）→小沙解题（4）→我通过眼睛看到了解题结果（5）

Then, by imitating the human thinking process, the operating steps of the unbounded system (2), (3), and (4) can be directly condensed.

Although the answer may not appear directly in your mind like you think about it yourself, it at least appears directly in your eyes.

Try to save as many intermediate steps as possible. The more steps you save, the closer the logic of the operating system will be to the human thinking process.

"The problem now has become more specific," Hao Cheng concluded, "that is, upgrading the recognition of the operational intent of the unbounded system to the recognition of the user's real needs.

“For example, my operating intention is to open the Xiaosha app and ask a question.

"The real needs of users are actually asking this question. If you pick up your phone, you will see the answer to this question, or our headphones or other external devices will directly play the answer to this question.

"So, isn't this one step closer to what we think? At least in terms of the situation, yes!

“As for how we know the user’s specific intentions: 1. Daily usage behavior; 2. Information received suddenly; 3. Things they want to do recently; 4. Others.

"Is this thing now possible again?"

It can indeed be done. Based on the following three points: First, Xiaosha knows users best, especially after using hardware fusion perception.

Second, as an all-round assistant, Xiaosha has a copy of all the user's information. After long-term use, Xiaosha's logic will tend to fit the users.

3. Xiaosha’s intelligence and ability are often not inferior to those of humans.

How can we not do it if we combine these three points?
This is the big data of AI. With Xiaosha's ability, he can really make it very accurate over time.

Moreover, this can not only help users solve problems, Xiaosha can also generate inspiration!

What is inspiration?

That's what came to mind on a whim!
At this time, who can say that the inspiration that comes out through the sand is not inspiration?
According to your own situation, give yourself a new inspiration every day, choose the reliable ones and study them in depth, and choose the unreliable ones and give up directly.

Over time, there will definitely be results!
What was originally a very sci-fi thing turned into a reality after Hao Cheng set the target, took a step closer to the target, and then analyzed the process.

A seemingly impossible task eventually turned into an AI training problem:
From operational intention recognition, it is upgraded to user real intention recognition by combining users' long-term usage data with instantaneous information input.

And for Baiju Technology, this is not a difficult task at all. At least the prospects are already visible.

Just like that, the next update and capability of the borderless system has been presented to Wei Shujie.

Although it is indeed far inferior to directly absorbing memories, wouldn't the already extremely powerful unbounded system become even more powerful?

……

"Hiss!" The more Wei Shujie thought about it, the more shocked he became. The more he thought about it, the more narrow-minded he felt. He couldn't help but sigh in his heart: With Baiju Technology, wouldn't the progress of the world be accelerated?

Today, he witnessed Hao Cheng's way of thinking and logic for the first time, and it was like this.

He thought of his own [Logical Operation Unit for Non-Deterministic Input] again. It seemed that the way this topic was generated was almost the same.

He turned to look at Hao Cheng, who was also looking at him and nodded, telling him clearly that it was.

Set up a target in the distance, that is the direction, and take a step towards it, that is progress.

"look,"

Hao Cheng also showed Wei Shujie his mobile phone, and Xiaosha summarized what Hao Cheng had just said.

That’s not all. Based on these, Xiaosha also made an update strategy for the official version of the Unbounded System, and even made specific plans and updates, and even prompted whether to send it to Lin Wen.

This was not something I should have discussed with Wei Shujie, but the conversation ended here.

"Look at this. If we upgrade it a little bit, wouldn't it be similar to the operation intention recognition I mentioned?" said Hao Cheng.

"Indeed!" Wei Shujie let out a long breath and felt a little short of oxygen - he really felt that he could not keep up with Hao Cheng's pace.

"What did you come here to talk about today?"

Hao Cheng also looked at Wei Shujie and asked, while authorizing Xiaosha to send the plan of the Boundless System 2.0 to Lin Wen.

"The problem of indeterminate input."

Wei Shujie's mouth twitched. He had just talked about a little bit of his achievements when he was distracted. He also helped Hao Cheng sort out such a big idea: "I decided to start with the judgment of the concept of 'more' or 'less'."

"Yeah!" Hao Cheng nodded and motioned Wei Shujie to continue.

"It is still very difficult for machines to judge whether it is more or less, and the way machines judge is very different from that of humans."

Wei Shujie continued, “For example, if there are a group of people on the street, whether they are many or not, humans only need to take a quick glance to draw the conclusion that they are many!

"It is very difficult for a machine to make this judgment. After collecting the image, the machine must first count, and then there must be a standard for how much is considered a lot.

“That’s a huge difference.”

"Indeed," Hao Cheng agreed with Wei Shujie's idea: "This is a very good direction. Do you have any specific ideas?"

Wei Shujie shook his head first, then half-nodded: "Let's start with Go."

There is a problem when AI plays Go: I occupy a large area, how much is this large area? People actually estimate by visual inspection, but AI can really calculate accurately.

Wei Shujie's plan is to change the algorithm of AI so that it can also estimate based on the size of the "slice".

"Have you ever done any specific hypothesis evaluation?" Hao Cheng had just looked at Wei Shujie's hypothesis evaluation of the overall uncertain state and found that he was indeed good at doing this.

It means first assuming that this thing can be realized, and then before it is realized, you can estimate the efficiency of this thing.

"Assuming that AI can roughly estimate the quantity through 'slices', when judging non-quantitative indicators such as more or less, good or bad, the theoretical speed increase is at least 10,000 times.

"Of course, the accuracy will be a little lower, 95%!"

Hao Cheng nodded and imagined a scenario in his mind: a Go board with two empty spaces, and he had to decide which one was bigger:
Today's AI is capable of precise counting, which requires tens of thousands of calculations.

If you make a judgment on whether it is more or less like a human, you only need to take a quick glance, which means only one or two calculations.

The speed is indeed thousands of times faster, but the accuracy is slightly lower.

Is it worth sacrificing 5% accuracy in exchange for a -fold speed increase?

That depends on the usage scenario.

Especially in AI training, this uncertainty not only has no effect, but can even bring more variations and possibilities.

"Do you have any specific algorithm ideas?" asked Hao Cheng.

"Yes!" The main purpose of Wei Shujie coming here today was to tell Hao Cheng that "yes"!
"Okay, let's do it this way first." Hao Cheng nodded.

It is not just a problem for machines to judge "more" and "less" like humans, but a class of problems. As long as it is really solved, its future applications will be very extensive.

Hao Cheng even believes that this is the key to whether AI can really go one step further - because he tried it, and Xiaosha's current way of judging more and less is still counting.

(End of this chapter)