The study of thinking models is still a process of squeezing out toothpaste.
It is necessary to constantly replace formulas, correct parameters, discover problems, and solve problems.
This is a process that requires patience, and it also requires huge computing power, not only the computing power of the computer, but also Chu Fei's own thinking ability.
In the virtual space, four mental models are calculated at the same time; Chu Fei also needs to observe the four models at the same time and constantly correct them.
This is also a research that requires a solid foundation of knowledge - what kind of formulas are needed, how to replace the formulas, what kind of parameters to substitute, etc., all require a wealth of knowledge.
Similarly, in such a research process, it is also a process of continuous practice, testing one's own knowledge system in practice.
With the calculation, data is constantly filtered out and entered into the database.
And with the accumulation of databases, empirical formulas have become more perfect, and even topological structures have become richer.
The entire empirical formula gradually formed a trunk and branches, and the branches gradually crossed to form a network.
A topological network gradually forms.
Only now did Chu Fei realize the changes:
"This is an intelligent neural network system, at least it can be regarded as a prototype of a neural network."
Discovering this, Chu Fei went back to look for related technologies and architectures of artificial intelligence, and continued to integrate them into calculations. At the same time, he also looked for similar AI structures for reference, and even excerpted some codes and embedded them into his own system.
The computing speed of the four thinking models has increased sharply, the database has accelerated and become richer, and the primary neural network system has accelerated its spread, and has begun to transform from a flat structure to a three-dimensional structure.
An AI structure is rapidly taking shape.
With the formation of AI, the calculation of mental models is in turn promoted.
The calculation of the mental model actually began to accelerate.
After two days of this, a new stable thinking model structure was established, and the computer spontaneously completed a verification work.
At this point, the structure of the fourth generation thinking model has been determined. This is a more detailed and sophisticated structure.
Within the thinking model, logical chains and topological structures are criss-crossed. Although it is essentially the core of Euclidean geometric space, large spaces are nested within small spaces, interlocking and embedded in each other.
The basic thinking model is essentially a tetrahedron structure. Each small tetrahedron is like a small computing unit. Countless tetrahedron units are continuously stacked, eventually forming a complex thinking structure.
But this time the computer's behavior is different from before: although the fourth generation has been determined, the computer is still running, autonomously calculating the fifth generation thinking architecture.
After taking a look at the AI network that was running spontaneously and taking shape, Chu Fei exited the virtual space, formatted his third-generation thinking model again, and began to build a fourth-generation thinking model.
It still has a dual-core and four-thread structure, but because the structure is more delicate, it took Chu Fei seven hours to build the mental model, consuming a super potion, four barrels of meat porridge, and a pack of nutritional supplements.
After opening his eyes, a flash of hope flashed in Chu Fei's eyes - this time the thinking structure should be better, after all, AI deduction is used!
Try it out!
After a few minutes, the score was run dozens of times, and the final score stabilized at: 14077 times/second!
Compared with the 11056 times/second of the third-generation thinking model, the computing power has increased by 27.3%!
Chu Fei wanted to cheer and scream.
What is exciting is not only the results of the fourth generation thinking architecture, but more importantly, the technology has begun to enter a state of accelerated iteration.
Chu Fei once knew this saying: There is an essential difference between the era of computer technology and the era of mechanical technology in the past.
In the age of mechanical technology, a technology requires more than ten years, or even decades of pondering and research to be perfected bit by bit.
But in the era of computer technology, computing power is king. As long as there is enough computing power, technology can be iterated quickly. Sometimes it can be iterated dozens of times in three to five years, and technology will be completely transformed.
Now, I seem to have one foot on the fast train of computer technology.
After completing the construction of the fourth generation thinking model, Chu Fei rested overnight, or rested for the rest of the night. The next morning he went out for a run, exercised, said hello to everyone, checked the inventory, and entered the virtual space again.
Check progress.
But it was discovered that in just one night, 22% of the deduction of the fifth-generation thinking architecture was completed!
In the thinking space, a progress bar actually appeared.
This progress bar is derived based on an empirical formula. Although it may not be accurate, it can still be used as a reference.
Chu Fei carefully checked the progress and reviewed the calculation process, and found that there were still some minor problems with the AI automatic calculation, but he only needed to manually adjust the parameters.
Then, continue to incorporate more scientific knowledge, mature model structures as references, and some experience summaries, etc.
There is still a small problem with the current AI. It does not actively search for relevant knowledge, even if it is in the database. Chu Fei still needs to input it manually.
In addition, as the amount of input data increases, computer efficiency begins to decrease. Previously, four mental models were deduced at the same time, but now due to limitations in computing power, one mental model had to be cut off, leaving only three.
Even so, it still only took five days to determine the fifth-generation thinking model architecture.
Compared with the fourth generation, the fifth generation is mainly an optimized and streamlined architecture.
We have solved some minor problems that could not be solved during the development of the fourth generation. We have replaced some additions with multiplications, replaced some complex number sets with combination curves, differentials, and probability formulas, or rearranged some chaotically distributed data.
, turned into small data packets.
Under the guidance of this streamlined thinking, the fifth-generation thinking architecture is much simpler than the fourth-generation architecture, but the computing speed has increased a lot.
Chu Fei took a break, formatted his thinking model again, and built a fifth-generation thinking model.
Because the fifth-generation architecture is streamlined on the basis of the fourth-generation, it only took Chu Fei four hours to build the fifth-generation architecture.
The final running score test: 15859 times/second, the computing power increased by 12.66%.
Feeling the continuous increase in computing power, Chu Fei was filled with satisfaction and even greater desire:
Is this the thrill of upgrading?
Not sure if we can continue to iterate?
But after logging into the virtual space, I found that the sixth generation thinking architecture was only 2% advanced!
"It has reached its limit!" Chu Fei understood. "Without external technology reference and more technological input, the sixth generation technology will be difficult to produce."
Sighing, Chu Fei became excited again. "The current thinking model is already very good. What will happen if such huge computing power is used to drive spells?
I want to rebuild the wind of perception, the butterfly transformation, and the refining magic weapon!"