**Star News, on February 28, 2024, Yanshan Technology (002195) announced that the company would accept institutional research on February 28, 2024, with the participation of people's livelihood ** Lv Wei Jin Yuxin and investors.
The details are as follows:
Q: What is Brain-inspired AI?
A: Brain-inspired AI is an artificial intelligence system that mimics the structure and function of neural networks in biological brains. Its design is inspired by the neuronal network of biological brains, which attempt to mimic the brain's ability to learn, perceive, and make decisions. Brain-inspired AI systems have the ability to perform complex pattern recognition, language processing, and decision-making tasks. Brain-inspired artificial intelligence is an important part of brain science research, involving medicine, biology, computer science, materials science, data science, social sciences and humanities, and its core content is the accurate analysis of biological brain neural networks. Brain science and artificial intelligence are the two hot spots of international scientific research today, and brain-inspired artificial intelligence is an emerging interdisciplinary discipline combining these two disciplines. Its goal is to use the latest brain science and artificial intelligence technologies and tools to decipher the structure and function of the biological brain, draw a map of brain function, structure and information processing, deepen the understanding of the working principle of the biological brain from the micro, mesoscopic and macroscopic levels, and build an artificial neural network system that simulates the biological brain, and finally achieve the goal of "understanding the brain, protecting the brain and simulating the brain".
Q: What is a brain-computer interface (BCI)?
A: Brain-computer interface (BCI) usually refers to the establishment of a new type of information exchange and control channel between the brain and the external environment without relying on conventional spinal or peripheral neuromuscular tissue systems, so as to realize direct interaction between the brain and external devices. In addition, brain-computer interfaces involve two-way interactions, which include feedback from a computer or the environment that can influence brain activity through neural modulation techniques. Brain-computer interface is a subdivision of brain-like artificial intelligence research, which aims to break the bottleneck of information interaction between the brain and the outside world, and is the only way to achieve human-computer interaction and human-computer integration.
The principle of BCI technology is to collect the activity signals of the brain nervous system through data acquisition equipment; Then, the recorded data is converted into a digital signal that can be recognized by the computer by amplification, filtering, analog-to-digital (D) conversion, etc.; Subsequently, the neural signal processor was used to preprocess the recorded signals, extract features, and then convert the features into output instructions to replace, restore, enhance and supplement brain functions. Q: What is a brain-computer interface system? How are companies in the industry progressing?
A: A typical BCI system mainly consists of four parts: signal acquisition, signal processing, equipment control and feedback. Among them, the signal processing part includes preprocessing, feature extraction, and neural decoding. The technical system of brain-computer interface is mainly divided into hardware layer and software layer. The hardware layer includes EEG acquisition equipment and EEG signal processing equipment. EEG acquisition equipment includes core materials, devices, and electrodes, and EEG signal processing equipment includes chips, power supplies, etc. The software layer includes EEG signal pre-processing and analysis, core decoding algorithms, communication computing, and security privacy. With the continuous progress of materials science, signal processing, and medical equipment, the amount of EEG signals that can be collected is getting larger and larger, and how to extract the required granularity information from massive data, among which the EEG decoding algorithm is the key to the urgent breakthrough in the brain-computer interface system.
In recent years, the brain-computer interface industry has developed rapidly, and Neuralink, founded by Musk, focuses on the research and development of brain-computer interface, explores the implantation of devices into the human brain to record brain activities, subverts the business model of traditional brain-computer interface companies, and develops R1 surgical robots and N1 brain-computer interface chips. Blackrock Neurotech introduced the NeuroPort electrode and pushed the Utah array into the mainstream of the BCI industry. Google's DeepMind Lab in the United States has outstanding research results in machine learning algorithms and artificial intelligence, which can efficiently classify and recognize brain signals and achieve precise control of brain-computer interfaces. The Braingate team in the United States has also made an important breakthrough in brain-computer interface data processing, using machine learning algorithms to achieve accurate recognition of motion intentions and limb movement control. Q: What are the subdivisions of the company's research in the field of brain-inspired artificial intelligence?
Answer: On the basis of inheriting the research results of the controlling shareholder in the field of brain-inspired artificial intelligence for many years, the company established the Yansi Brain-inspired Artificial Intelligence Research Institute in August 2023.
At present, the Institute of Brain-like Research is mainly committed to the research of cutting-edge fields such as (1) the analysis and regulation of the internal state of the brain, (2) the deep generative brain signal decoding algorithm (also known as the "EEG model"), and (3) the diagnosis and intervention of non-organic major brain diseases. Q: Why does the Institute focus on the analysis and regulation of the internal state of the brain, as well as the research of the large EEG model?
Answer: The analysis and regulation of the internal state of the brain is the key to understanding and protecting the brain, which is of great significance for human beings to understand the biological brain and diagnose and treat major brain diseases. The establishment of EEG large model is the core technology to achieve real-time, accurate and multi-dimensional neural decoding, which is recognized by the industry as the focus and difficulty of research in the field of BCI.
Just as massive corpus is a necessary condition for large language models, the fuel of brain-like artificial intelligence is massive EEG data. With the continuous technological progress of BCI hardware manufacturers such as Neuralink, BlackrockNeurotech, Braingate, and Synchron, invasive hardware will eventually reach a more advanced state than it is now, so as to obtain high-throughput brain neural activity data of patients or normal people more safely, efficiently, accurately and conveniently, and the collected EEG signals will increase exponentially. How to interpret the brain's intention in the massive EEG signals will become the bottleneck of human-computer interaction. Just as the current language model has achieved great success, the construction of EEG model in the future is an inevitable choice for brain-computer interface and human-computer interaction. Based on the above thinking, Yansi Brain-like temporarily skips the research and development of electrodes, chips and other hardware, and directly lays out the construction and research and development of EEG large models in advance, so that it can adapt to the massive EEG neural network data obtained in various ways such as non-invasive and invasive now and in the future, and empower the hardware with the EEG large model, so as to achieve a real-time, accurate and efficient human-computer interaction system. Q: What is an EEG model? What is the research goal of the EEG model?
Answer: The EEG model is a super-large deep learning model pre-trained based on massive brain neural network activity data, which learns the intrinsic expression and dynamic characteristics of brain nerve signals through pre-training, and has the generalization ability to analyze various complex functions of the biological brain. As the underlying algorithm model, the EEG model empowers cutting-edge fields such as brain science, brain health, brain-computer interface, and human-computer interaction.
The industry-recognized research goals of EEG large models include learning, understanding and simulating the operation of biological brains to the greatest extent, creating an I that thinks like a living thing or even a human (i.e., "ChatGPT" of brain-like artificial intelligence), and then translating the ideas in the brain and outputting them to downstream peripherals (such as robotic arms, humanoid robots, etc.) or downstream i systems (GPT, Wensheng**i, etc.) to achieve real-time, high-throughput unbounded communication between the brain and the external physical world or the virtual metaverse. Q: What is the company's research method in the field of EEG large models? What is the current stage of the research on the EEG model of Rock Brain?
Answer: The Institute of Brain-like Research is constantly trying to solve the problems of brain science by combining i and brain science, and realizes the research and development of large EEG models by using generative I and contrastive learning, as well as with professional brain science experimental paradigms. The R&D process includes iterative steps such as data collection, data cleaning, preprocessing, data tokenization, model pre-training, fine-tuning based on downstream tasks, model validation and optimization, etc.
At present, Yansi Brain-like has begun to try to pre-train the EEG model, with the aim of making the EEG model learn the change law of EEG signals, and then use the compressed intrinsic features of the underlying layer of the model as the input of downstream tasks, so as to complete the translation of brain ideas. In the future, when the EEG model matures, it can be connected to peripherals to complete the interaction of brain-computer interface or the metaverse, or cooperate with other multi-modal large models to realize the concrete display and real-time interaction of brain ideas. In addition, the EEG model can also be used for early screening, intervention and efficacy evaluation of non-organic brain diseases by analyzing the abnormal state of brain activity of test subjects. Q: What is the progress of the commercialization of RockSys brain? What areas can the research results be used for?
A: The EEG model of the Yansi Brain-like Research Institute is still in the research stage and has not yet generated operating income. Theoretically, the research results can be applied to early screening and intervention of non-organic brain diseases, brain science research, human-computer interaction, intelligent driving, robotics, metaverse and other fields.
The main business of Rock Mountain Technology (002195) is Internet information service business, artificial intelligence business and diversified investment business.
According to the third quarter report of Rock Mountain Technology in 2023, the company's main revenue is 43.1 billion yuan, down 1508%；Net profit attributable to the parent company 32.9 billion yuan, up 1. year-on-year19%；Deduct non-net profit 31.3 billion yuan, an increase of 362%；Among them, in the third quarter of 2023, the company's single-quarter main revenue is 13.8 billion yuan, down 986%；The net profit attributable to the parent company in a single quarter was 6952460,000 yuan, a year-on-year decrease of 3185%；The non-net profit deducted in a single quarter was 582160,000 yuan, a year-on-year decrease of 3977%；Debt ratio 308%, investment income 11.8 billion yuan, financial expenses - 15.3 billion yuan, gross profit margin of 5528%。
There are no agency ratings on the stock in the last 90 days. Margin data shows that the stock has a net financing outflow of 23.7 billion, with a decrease in financing balances; The net outflow of securities lending was 2820750,000, the balance of securities lending and borrowing decreased.
The above content is compiled by **Star based on public information, generated by algorithms, and has nothing to do with the position of this site. **Star strives but does not guarantee the accuracy, completeness, validity, timeliness of all or part of the content of the information (including but not limited to text, **audio, data and charts), if there is any problem, please contact us. This article is a compilation of data and does not constitute any investment advice for you.