Reborn in 2010, the AI ​​in my mind is too powerful.

The SEELE R&D platform is currently only version 1.0, but there's not much room for improvement.

Thanks to Jasmine's underlying optimization, the software aspects can be easily improved by simply adding layers on top of it.

The main issue is the hardware; it's not keeping up yet.

Ideally, numerous subsidiaries would be connected to a more intelligent system, gaining more intelligent computing power, leveraging the power of AI to dominate the market, with lunar exploration in 3 years and Mars exploration in 5 years just around the corner.

However, in reality, the current hardware, especially in chip manufacturing, is really quite inferior.

Although companies have praised SEELE after using it, they generally report that the AI ​​computing time is insufficient.

Several branch companies have already begun to compete with each other, sending messages and whispering to Chen Yiran, each one like a peacock spreading its tail feathers, explaining the importance of their company in the group's future plans, and hoping that the R&D platform can give them some preferential treatment.

Chen Yiran ignored them completely. After actually visiting the wafer fab and seeing the entire chip manufacturing process, he felt much calmer.

Bottlenecks in the manufacturing industry are an objective reality.

Even though he brought many new technologies, it would be quite difficult to actually double the speed of technological upgrades.

Chen Yiran could only start making arrangements in other areas.

The most noteworthy issue is brain-computer interfaces.

Since last year, Chen Yiran has been secretly following the development of brain-computer interfaces.

The current development of brain-computer interfaces is quite different from what he remembered.

In his previous life, he had only seen news about brain-computer interfaces in the news.

As a layman with limited knowledge, he always thought that the high-tech brain-computer interface was like the one in science fiction movies, where someone wears a small hat and then wreaks havoc.

However, in reality, this is completely impossible. The hard-implementation design found in cyberpunk is currently the only feasible approach.

Because the electrical conductivity of brain tissue itself is about 0.3 S/m, while that of cerebrospinal fluid is slightly higher.

However, the skull is a dense bone plate with very low electrical conductivity, only 0.01 to 0.04 S/m.

So after the signal comes out of the neuron, the skull acts as an electrical deceleration barrier in the middle.

When a neuron fires, a tiny electric field is generated around it. If an electrode is placed next to this neuron, a clear and distinguishable discharge spike can be measured, which is about a few hundred microvolts, lasts for one millisecond, and has a unique waveform.

However, when this electric field passes through the skull, the current does not flow straight towards the top of the head. Because the resistance of the skull is twenty to thirty times higher than that of brain tissue, the current is forced to spread out on the inner surface of the skull, finding the path of least resistance to pass through the sutures and blood vessel openings.

When it reaches the scalp surface, what was originally a focused point of discharge from a single neuron will become a blurry potential fluctuation with a diameter of about three to five centimeters.

The superposition of hundreds or even thousands of firing neurons, after passing through the skull, will be spatially low-pass filtered into a blurry waveform.

Jasmine had carefully calculated that even if AI computing power made great strides and accumulated enough computational data, AI noise reduction and decoding would still be possible.

If only electrodes are used to collect brain signals on the scalp, it can only decode coarser signals.

This is limited by physical properties, according to Shannon's theorem.

Given the bandwidth and signal-to-noise ratio of a channel, the maximum information rate you can transmit is limited. Exceeding this limit will result in information loss regardless of the encoding method used.

Therefore, the mainstream approach to brain-computer interface research currently involves implantation into the brain.

There is a groove on the surface of the brain called the central sulcus, and the ridge in front of the sulcus is called the precentral gyrus, which is the primary motor cortex M1.

The M1 has a body map with the feet at the top, the hands in the middle, and the face at the bottom. The area for the hands is called the hand area, which is much larger than the leg area because fine motor skills in the hands require far more neural resources than in other parts of the body.

This is the main direction of current research, because even if a paralyzed person tries to move their hand, the neurons in M1 will still fire.

Regardless of the reason for the disruption of the pathway from M1 to the muscle, the neurons within M1 themselves are not dead. They are still alive and continuing to send signals according to their original coding patterns.

Most organizations use the Utah Array for implantation testing. This concept, proposed by Professor Richard Normann of the University of Utah in 1989, is a microelectrode array implanted in the cerebral cortex.

Essentially, it is a square grid consisting of about 100 silicon needles, each with a tiny electrode contact at its tip.

100根硅针组成10×10网格，针距 400微米，裸硅尖端 1–5微米，镀铂后 10–20微米，

These silicon needles are made using micromachining technology. The needle spacing is about 400 micrometers, the bare silicon tip is 1 to 5 micrometers, and even after platinum plating, it is only 15 micrometers, which is thinner than a hair. The insertion depth of 1.5mm is just enough to penetrate the surface layer of M1 gray matter.

The entire array is designed for out-of-plane insertion, allowing it to be sealed within the skull after implantation, making it suitable for long-term recording.

Over the past decade or so, quite a few institutions have conducted research in this area.

BrainGate is the organization that has gone the furthest.

This is an academic clinical trial consortium led by Professor John Donohue of Brown University, with the participation of Massachusetts General Hospital, the U.S. Veterans Affairs Medical Center, and the German Aerospace Center.

The first human implantation trial was completed in 2004, using the Utah array.

The Utah array also became the world's first high-channel microelectrode platform to receive FDA approval for implantation in the human cerebral cortex.

2012年5月17日《自然》论文发表了他们2011年4月和10月做的两次实验。

The subject of the experiment was named Cathy Hutchinson, who had been paralyzed for 15 years due to a brainstem stroke and was currently quadriplegic and aphasic.

The Utah Array was implanted in November 2005 and has been recording data ever since.

On April 14, 2011, she used a 96-channel Utah array implanted in her motor cortex to control a robotic arm to pick up a coffee bottle and take a sip.

This was the first time in fifteen years that she had completed the act of drinking by herself.

Another subject in the experiment, Bob, was a 66-year-old man who suffered a brainstem stroke in 2006 and received a Utah array implant in June 2011.

In October 2011, a robotic arm was able to grasp foam balls. This action was more complicated than drinking water because the ball moved slowly, but it still achieved a success rate of 62%.

These two experiments demonstrated the persistence of the Utah array, which was able to stably record neural signals even five and a half years after implantation, a lifespan that far exceeded previous expectations.

This also proves that this implantation does not require any additional training and can be used one or two weeks after implantation.

From the perspective of human development and medical advancements alone, these achievements are quite remarkable.

But for Chen Yiran, these were simply not enough.

He developed brain-computer interfaces not to benefit the public, but to deconstruct his own Jasmine.

So, after consulting a lot of information, he turned his attention to another large project.

That is the Human Connectome Project, launched by the U.S. federal government in 2009, and the MEG magnetoencephalography (MEG) technology, which seems to have no practical value but can solve some pain points.