How Decentralized Artificial Intelligence And Machine Learning Change Medicine
Artificial intelligence (AI) is as a specific type of software capable to accomplish activities like a human being. This includes visual perception, voice recognition, making decisions, cross-language translation, etc.
Machine learning (ML) is commonly considered as a part of AI. It includes prognosis tasks, taxonomical classification and other processes generating relationships. The core part is getting new data from different sources as it was programmed.
There are many ML and AI levels exist, including:
- Supervised learning
- Unsupervised learning
- Deep learning
A huge data volume must be consumed at all levels to generate reliable internal connections between key components in a proposed knowledge area. The might of AI and ML allows even to overcome human abilities in the learning sphere. Having enough time and sufficient complexity, AI can simulate cooperative behavior and follow a personal data security attitude.
The very same features are inherent to decentralized AI. With ML support, all data will be safe and protected. The main reason for this is the way data is managed and transferred among participants. After all, the learning entity is “growing up” and allows everyone across the network to access its achievements. This removes the necessity of a single governing institute. Importance here is hidden in decisions taken by an authority alone, sometimes in contradiction to the society opinions.
Blockchain technology improves cooperation
On-chain cooperation depends on agreed patterns. They describe an assets transit or other aspects of participants’ behavior with the help of smart contracts (smarts). The smarts are places in blockchain in each hosting node with the ability to be verified by them separately. The union of decentralized peers compares requests to the smarts and act in correspondence. A single smart is dedicated to a single function, describing it the way its authors want to. They need no trust among parties at all. A big group of smarts can even form a kind of an interactive system with open access. Everyone could join this project without any centralization expenses.
Today, mobile phones and tablets are widely used by people. Their total computing power is impressive. All those devices constantly receive data that usually has a complex private nature and a centralized storage facility, thus they must effectively protect it. Consequently, all methods used to receive information for training can participate in the usability and security improvement.
There are a few good examples. Considering medical solutions, some startups like Neuron are worth mentioning even in their beta state. They are dedicated to decentralized AI training procedures explained to users (or paraphrasing, how to teach the teacher). Participants will find out how to create a dataset of their health parameters and how to correctly use it.
Computer Vision Upgrade
A simple photo can be a source of your health dataset, if AI and ML are applied in a proper way. This is the main goal of one of the mentioned projects. It is called Selfie2BMI and uses Deep Neural Networks with optimized tangible anatomy details prediction including height, weight, age, and gender. Talking only about facial characteristics, there are more than 23 parameters monitored, including color, teeth, eyes, etc.
Analyzing Blood Type
One more Neuron project is dedicated to expanding blood test results by polling. It is planned to be performed by an AI agent with 400 blood parameters taken into consideration during the dialogue. Several thousands of medical common FAQ were taken to train it along with other medical documents. The agent is able to distinguish users in accordance with their age, gender and other initial parameters to improve the result quality.
Another dialogue agent is developed to empower genetic advisory with the help of multiple questions concerning various educational and personal aspects. It records and keeps all the recommendations it has ever done to a visitor and can use a huge database to search from.
A simple module, which allow choosing the dose of medicine taking into account all side effects and previous medical history, is developed as well. It also uses pharmacogenomic data if a correspondent module is attached.
This approach get a user closer to the medical treatment.
Unfortunately, some problems with decentralized AI still exist preventing user to manage its own medical history.
The Burden of Integration
All proposed projects are intended to support users in searching for their medical parameters. The majority of people has no access to medical data. Moreover, they have no idea where to start from and what to do. The open source nature of innovation company frightens people as well. Such a lack of support stops data decentralized integration from being implemented at will.
Blockchain here plays a role of a limited accessible account carrier with a full information sources supervision and verification. This brings up a huge forecast potential along with the opportunity to check data source in accordance with the Know Your Data principles. As for the data dependence issue, the proposed solution brings more effective healthcare for participants.
Users may reject sharing their medical information over the network because there is a risk of unwanted publication. Blockchain uses a hard cryptography approach for information being kept safe and at the same time available. Moreover, Neuron corresponds to HIPAA requirements by storing data on a specific local device.
The invention of a machinery doctor is still far from being reached, even though the industry is rapidly developing. Doctors are commonly divided into generalists that might be described as hardly developed general AI, and specialists which are much easier to model. ABMS (American Board of Medical Specialties) listing shows that there are over 150 specialties and subspecialties in the medical field. This is where Neuron can find its main opportunities.