茶杯狐 -CUPFOX

As an IT professional who always keeps up to date with the latest information, Xiang Kun is naturally aware of what the most advanced and cutting-edge AI looks like.

In the AI field, deep learning and neural networks are still dominant, with large factories at home and abroad, such as Doge, Penguin, Du Niang, Ali Niang, and so on, each having their own deep research direction, and they publish their latest progress from time to time.

After all, most of the current AI research directions are difficult to create good results with isolated work. Especially in machine learning, data, computing power, and algorithms are all necessary. Without a sufficient dataset, even if your algorithm is exquisite and your machine’s computing power is advanced, it’s hard to get the results you want and apply them to reality.

In some specific fields, AI has already surpassed, or even outstripped, human abilities.

For example, AlphaGo has defeated the top professional Go players one after another. With further improvements and iterations, human players may find it hard to compete with it in the future, and its opponents will become other AIs.

But can we really say that the players defeated by AlphaGo, such as Ke Jie, were only defeated by the AI, by a machine?

Perhaps it’s not that simple.

For example, Xiao Ming’s fighting capacity is only 5. He wants to meet his girlfriend, Xiao Hong, but on the way, he’s blocked by a dragon with a fighting capacity of 100; he can’t beat it.

So Xiao Ming went back to the village, gathered his dwarf friends, spent several years studying the dragon’s weaknesses and combat methods, conducted various simulated confrontations, and eventually built a flying sword, the “Bai Emperor Sacred Sword”.

The “Bai Emperor Sacred Sword” has a combat power of 200 when fighting dragons!

So, Xiao Ming, shouting “Bai Emperor Sacred Sword, follow me!” easily defeated the dragon and met Xiao Hong.

So, was the dragon defeated by the “Bai Emperor Sacred Sword”, or by Xiao Ming and his dwarf friends?

Without Xiao Ming, the “Bai Emperor Sacred Sword” might be knocked down by a cat with a fighting power of 3 because it had never seen one in its training set~!

Last year, Doge’s BERT model even achieved STOA results in 11 NLP tasks in the top-level machine reading comprehension test SQuAD1.1, surpassing human comprehension. Subsequently, domestic teams like Brother Xunfei, Penguin, and Ali Niang, have all topped the SQuAD test one after another.

In the SQuAD2.0, teams like the Asia Institute of Microsoft and Ali Niang are continuously achieving better results. It looks like humans have already been left far behind by AI in reading comprehension.

But is that really the case? Does AI truly have a stronger reading understanding capacity than humans? Can it understand human thinking and intentions?

After all, the current AI is still just a “high-level tool”.

No matter how human it appears.

When Xiang Kun first started the “AI Assistant Plan”, he had already seen the bottlenecks in current machine learning and neural networks. So in the beginning, he did not have the idea of creating a “super AI” because it was unrealistic. He had always been thinking about integrating and optimizing within the existing frameworks.

However, he later found that such a method would still take a lot of time and energy, and it was not something he could easily accomplish on his own. But since it required so much effort, why not think about taking a step further?

To put it bluntly, after several mutations, he has “expanded” again, holding two cents, he wanted to eat the best wonton in the world.

Because he thinks his two cents are very valuable, buying something worth a normal two cents, or even two dollars or twenty dollars would be a loss.

But how can he do this?

Xiang Kun has had several moments of “flash of inspiration”, and has a vague idea, but can’t yet fully grasp it, it’s still not ready to “co-manifest”.

That’s why he agreed to accept Zhang Qian’s project. He wants to use the process of doing other projects to help him find “that light”.

Unfortunately, the help had not been significant. The framework of that project was already set up and the core was completed, but Xiang Kun’s “AI Assistant Plan” still hadn’t even found a good angle to dig from.

Looking at the diagram on the screen, which he drew to help himself think, Xiang Kun pinched his chin with one hand, and played with his Carved Alsace with the other hand.

Xiang Kun now understands:

To make something worth two million with two cents, there must be highly innovative creation. It has to be something completely new, starting from the underlying level, a “lever” that can move the earth.

In the existing framework, there’s no way he could achieve that, even though his brain is much stronger than the average person, or even an ordinary genius, due to mutation. All the institutions, factories, talents, and teams in the world are studying every day, with so many devices and data for them to use, he can never catch up on his own. Since they can’t find the breakthrough from the old framework, it’s unlikely that he could do it on his own.

But how deep would this underlying level need to be?

If you look at it, innovations in software can never be separated from hardware.

Just like when the computer was first born, artificial intelligence was divided into “symbolism”, “connectionism”, and later “behaviorism”.

“Connectionism,” also known as the bionic school, advocates imitating human neurons, using the connection mechanism of neural networks to achieve artificial intelligence. The currently popular machine learning is carried out in the name of “connectionism”.

However, in the 1970s, due to the level of hardware development and lack of computing power, “connectionism” fell into a low point.

Until now, with the rapid development of computer computing power and the advent of massive big data brought about by the development of the Internet and mobile networks, machine learning, under the name of “connectionism,” has found the best growth environment and begun to flourish. It has spawned a variety of platforms, algorithms, and models and has seen a lot of applications in various industries and environments. There is even talk that “AI will replace human work”.

Earlier this year, Huawei released the Shengteng 910 AI training chip and the Shengteng 310 AI inference chip, and not long after, Ali’s Damo Academy also released the so-called world’s most powerful Hanguang 800 AI inference chip. These advances in hardware will undoubtedly lead to a more extreme evolution of machine learning.

In the foreseeable future, perhaps in twenty or thirty years, quantum computing will make significant breakthroughs, and quantum computers will begin to be used on a large scale. This will inevitably lead to major disruptions in the AI field, and might result in the introduction of a more extensive, complex, and robust AI model. We might even see a neural network that “copies” human neurons more thoroughly with the advancement of research, and thus, realize the actual meaning of “super AI”.

However, Xiang Kun can’t wait until then. Indeed, he doesn’t have enough money, or time, or the inclination to invest in hardware development and research. He’s not trying to lay the groundwork for the future. All he needs is an “AI Assistant” that can immediately lend him a hand.

Therefore, his “innovation” and his “foundation” can only be established based on the existing hardware conditions.

It may seem like he is daydreaming, but Xiang Kun genuinely believes that he is onto something and is only an inch away from a breakthrough that will shed light on everything.

Xiang Kun checks the time and notices it’s not too late. So, he picks up his phone and sends a message to Xiao Pingguo via WeChat: “Any plans for this weekend? Let your dad bring you over, I’ll cook something yummy for you guys and we can take ‘Glitter’ out to play.”

“That sounds great! But a visit won’t interrupt your schedule, right, Uncle Xiang?” Xiao Pingguo asked in an evidently delighted tone, though tinged with concern.

Xiang Kun replied, “No, it won’t be a bother. I’m a freelancer now; I make my own hours.”

After finalizing the time of their visit, Xiang Kun casually asked, “By the way, Xiao Pingguo, are you interested in computers?”

“Computers? I only know very little. During my time at the blind school, I took some computer-related courses, but I don’t really understand multiplexes, or any of those details,” Xiao Pingguo confessed.

“How about I teach you? You know, the domain of computers, especially AI, is extensive these days. One day, there could be AI specifically designed for ‘Bird Language’ interpretation, which could help you understand animal behavior. Maybe you could take a minute to learn about it. You might find it interesting,” Xiang Kun continued to guide.

“But I don’t have any foundation. Perhaps I should teach myself first…”

“No need. I can teach you from scratch. Just ask me whenever you’re unsure,” Xiang Kun firmly stated.

Then he started elucidating computer knowledge from the hardware level to Xiao Pingguo in simple language. This even included explaining how CPUs work, what the principles are, and how semiconductors come together to form gate circuits.

Needless to say, Xiao Pingguo was indeed a good learner. After Xiang Kun mentioned the potential future applications of AI in ornithology and zoology, she became somewhat interested, which made her willing to learn.

Furthermore, Xiao Pingguo was highly intelligent. Even though Xiang Kun’s explanations were simplified as much as possible, without real objects, the topic was still abstract. Despite this, she was able to understand it all quickly. When she came across something unclear, she only had to ask a few questions and would immediately have it figured out.

Nearing eleven o’clock, once he sensed a hint of fatigue and sleepiness in Xiao Pingguo’s voice, Xiang Kun actively ended the “lesson,” assuring her that she could leave him messages on WeChat if she had any queries and they could continue the next day.

After wrapping up the call with Xiao Pingguo, Xiang Kun started carving out some pieces from the pile of boxwood at home—he had replenished his stock with more rabbit and Boxwood from an old supplier last week when he was out.

Now these pieces of wood came in handy. Xiang Kun planned to carve them into computer components and then use the “interest” and “concentration” emotional modes to carry out “Emotional Infusions.”

By the time they visited this weekend, he would have just completed his Blood-drinking Period, and the effects of the “Emotionally Infused” objects would be ready to kick in.

During this period, Xiang Kun had been sensing the “Minion Wood carving” at Xiao Pingguo’s place every two days, guiding her dreams, and then “observing” in the dreams.

In the beginning, when he left Xiao Pingguo’s Dreamland, he could only establish cognitive models for objects that had appeared in Xiao Pingguo’s dreams or that Xiao Pingguo had seen or heard about using her special cognitive model.

However, after two Blood-drinking sessions and two rounds of mutation evolution, this cognitive mode seemed to naturally become one of Xiang Kun’s capabilities.

It was as if he had initially only installed software that couldn’t run because of hardware conflicts on his device, but the software could operate in the server—Xiao Pingguo’s Dreamland. Now, after the upgrade, his “hardware” has also begun to adapt.

Admittedly, compared with Xiao Pingguo’s cognitive mode, it’s still challenging for him to establish that special cognitive mode for all objects. He can only generally feel it, and the consciousness is intertwined with his original understanding, which is a somewhat disordered state.

Xiang Kun’s “AI Assistant Plan” was inspired not only by his mutation progress and theories but also by the unique cognitive models obtained from Xiao Pingguo’s Dreamland.

He had a vague idea of combining these two “things,” but he couldn’t find a suitable entry point. After all, both his understanding of his mutation and the special cognitive model he got from Xiao Pingguo were in a state of uncertainty—they worked sometimes and sometimes not.

Therefore, now he wants to guide Xiao Pingguo to learn about computers and then establish a cognitive model for the operational mode of the computer using her unique cognitive method. Xiang Kun would then use Xiao Pingguo’s Dreamland to perceive this model and see if he could gain any insights.

Instead of immersing himself in rumination, Xiang Kun thinks this method may be more promising.

Moreover, he genuinely adores Xiao Pingguo. If he could teach her some computer knowledge, it might prove to be useful in the future—after all, this is the Information Age.

Over the weekend, after enjoying a lavish meal prepared by Xiang Kun at his house and playing with Glitter, being enthusiastically tutored in computer knowledge for a few hours, and receiving a bunch of “audio teaching materials” recorded by Xiang Kun himself, Xiao Pingguo returned home with her father Li Yang and three unique woodcarvings.

“The dishes Uncle Xiang cooked tasted even better than last time!” Xiao Pingguo reminisced with a hint of regret: “It’s a shame that Uncle Xiang doesn’t live in the same city as us; if he did, we could eat his food frequently, hehe.”

“I never realized you were this food-driven before,” Li Yang chuckled, looking at his daughter.

“Who doesn’t love delicious food?” Xiao Pingguo replied, carefully touching the wooden box in her hand: “And these wood carvings are amazing too. Uncle Xiang managed to carve out the tiniest ‘capacitors’ and the most delicate circuits. He’s really talented.”

In the wooden box that Xiang Kun gave her, there were three wood carvings of different sizes–scaled-down carvings of a motherboard, CPU, and memory. It wasn’t that Xiang Kun was unwilling to carve out 1:1 sized versions, but that he couldn’t find a large enough piece of wood to do so.

Previously, when Xiao Pingguo was at Xiang Kun’s house, Xiang Kun deliberately disassembled a computer that he had bought for a project. Guiding Xiao Pingguo, he let her carefully touch the motherboard, CPU, and other internal parts of the computer. He detailed parts she was touching, their functions, and how it connected with what he had taught her via WeChat.

Seeing the wooden box in Xiao Pingguo’s hand, Li Yang also sighed in admiration: “Xiang Kun truly is a good person. I can feel that he sincerely wants to teach you about computers.”