As Gao Nian spoke, two professional terms appeared on the projection screen behind him.
They were 'ANN Artificial Neural Network' and 'SNN Spiking Neural Network'.
"There are many research and development directions for current artificial intelligence technology, each with its own characteristics. Among them, the neural network artificial intelligence technology route has two major technical branches.
And among these two major technical branches, the SNIN spiking neural network technology route is the closest to the human brain structure and operation mode.
It is the most power-saving and lowest power consumption technology route. From a long-term perspective, it has the most promising future. You can understand this by looking at the human brain's average power consumption of 20W.
Among them, our Geek Technology Company, Apple Company, and Google Company are all following the ANIN artificial neural network technology route.
As for why I want to talk about the 'SNIN Spiking Neural Network' technology route?
This is because our Geek NPU neural network processor can simultaneously run 'SNN Spiking Neural Network and ANN Convolutional/Recurrent Neural Network'."
As for what is the function of 'SNN Spiking Neural Network'? This depends on the operating principle of the human brain.
Our body perceives the world through the five senses, and then reacts to the external environment in real-time and quickly, which is very efficient.
Among these, has everyone noticed that when our eyes see something, we will definitely look at moving things first, and then look at stationary things?
Hearing Gao Nian's words, the people at the scene, including the netizens in the live broadcast room on the Internet, couldn't help but recall it, and found that the fact was indeed the case.
Compared to moving things and stationary things, their first reaction is definitely to focus their attention on moving things first.
Looking at the people nodding slightly, Gao Nian continued to introduce:
"The reason why this is the case is actually because our brain's operating mode uses the SNN spiking neural network operating mode.
Our brain has many neurons, and they have neural synapses. When a neural synapse receives a signal that meets the requirements.
At this time, the neuron will transmit the signal to the next neuron, thus forming a conditional transmission spiking neural network.
Therefore, although our brain has 86 billion neurons, its theoretical computing power is very strong, far surpassing the world's strongest supercomputer.
But because not all neurons work together at the same time, but a part of the neurons work conditionally.
Ultimately, our brain's power consumption is only a mere 20W, the power consumption is surprisingly low, and the efficiency is surprisingly high."
Hearing Gao Nian's words, the people at the scene also nodded slightly to express their approval.
They had actually learned about what Gao Nian said before, and knew how strong the brain's computing power was in theory.
They had also heard that the brain's power consumption was only about 20W, but it could do so many tasks, which showed how magical the brain was.
In response, Gao Nian continued while the people nodded slightly:
"The advantage of the SNN spiking neural network is its ultra-low power consumption. The power consumption is so low that it can be ignored when it is not working, and there is only power consumption when it is working.
This is simply too important for our smart phones and other mobile devices, as well as artificial intelligence.
Although the SNN spiking neural network naturally has the disadvantage that the data obtained is not as accurate as the ANN artificial neural network.
But when we deal with many problems, such as real-time processing, anomaly detection, or pattern recognition, etc.
In fact, it does not need to be so accurate, so perfect, as long as a general range of results is enough.
Therefore, when we process certain specific tasks, we can hand them over to the SNIN neural network to process,
and some tasks that require accuracy must be handed over to the ANN artificial neural network to process."
Hearing Gao Nian's words, the people at the scene nodded slightly as if they understood, but in reality, they still didn't know what Gao Nian was talking about.
However, what Gao Nian said next quickly made them understand the power of Geek Technology's next-generation artificial intelligence technology.
"How powerful is our Geek Technology's artificial intelligence technology? Everyone must have a deep understanding of this. It is no exaggeration to say that our artificial intelligence Xiaoke's scores in the evaluations of various institutions far surpass its peers.
And today, our Geek Technology's artificial intelligence technology has ushered in its second innovation.
This second innovation can currently only be enjoyed by Geek S3 smart phones, Xiaomi M2 smart phones, and Geek H1 tablets.
Because the second-generation artificial intelligence Xiaoke is built on the Geek NPU neural network processor that can simultaneously run 'SNN Spiking Neural Network and ANN Convolutional/Recurrent Neural Network'.
Without the Geek NPU neural network processor, there is no way to run the second-generation artificial intelligence Xiaoke.
Specifically, what is the difference between the second-generation artificial intelligence Xiaoke and the first generation?
The characteristics it exhibits are faster task processing speed, smarter, can do more things, especially in image processing, it can be called magical!"
After a pause, Gao Nian continued:
"First of all, the first reason for faster processing speed is because we can run neural networks at the same time, give full play to the advantages of the two networks, and the speed of processing problems is naturally faster.
The second reason for being smarter is because we have upgraded the artificial intelligence Xiaoke's program algorithm, which is naturally smarter than before.
The third is in the area of graphics processing. Our Geek NPU neural network processor supports the SNIN spiking network.
This means that our artificial intelligence Xiaoke can quickly screen dynamic image information from a large amount of information, and ignore or process non-dynamic image information later.
This forms a new brain-like visual perception model that imitates the human brain's dual pathways, and its ability to process dynamic information is very strong."
After a pause, looking at the confused people, Gao Nian continued:
"Specifically, what are the benefits when everyone actually uses it?
First of all, in terms of mobile phone photography, because our artificial intelligence Xiaoke is particularly good at processing dynamic image data, especially dynamic image data with ultra-fast processing speed.
This also means that when we take pictures, we are very good at capturing the moment of an object's movement, so that there is no visual afterimage in that moment of taking pictures!
Therefore, users will find that when using our mobile phone to take pictures, there will be no afterimage phenomenon.
It can be said that this is the strongest supplement to optical image stabilization technology. One is physical and the other is software, and the two form a perfect complement.
After that, there is video shooting. Because the SNN spiking neural network will first pay attention to 'intensely reacting' things, that is, the focus is on moving things.
Stationary things will not be paid attention to and ignored. The final result is that we can generate ultra-high frame rate images out of thin air, such as increasing the frame rate of recorded videos to one thousand frames or even ten thousand frames, thereby turning a blinking action,
into countless times of slow-motion effect.
Therefore, it is actually a slow-motion camera that can slow down the speed of all things moving!"
