Introduction of XAIN and its Technical Vision
We envision in a world, where equally distributed intelligent systems enable meaningful collaboration and trust.
This means that we focus on developing artificial intelligence systems that are directly applied on local data in a distributed system without the requirement to collect and aggregate all data into a central global database to then train the system but instead to achieve local intelligence to enable trust and to cover new scenarios in which classical machine learning could not be applied before, due to both technical restrictions, such as unreliable networks or legal restrictions, where individuals or organizations are not able to share data, e.g., accounting data.
Now, in order to achieve distributed machine learning, we require a reliable network that prevents data manipulation and provides an efficient verification basis share only the learning of different intelligent systems instead of the actual data behind the individual applications to gain network effects.
One of our scenarios hereby is, e.g., autonomous driving, which we developed together with the University of Oxford and Porsche as a leading German carmaker. In this case, some carmakers have a dominant position in centralized autonomous driving with over-the-air (OTA) updates, by collecting sensor data of test cars, enriched by synthetic data, with a focus on markets, such as the US, which constantly results in issues in European and Asian driving conditions. As such, it is a global solution to very local problems. Using distributed machine learning, the vehicle is trained to drive on the basis of local data, enriched by synthetic and general data, such as weather conditions. Now, to achieve network effects, also in situations of low network bandwidths, we only share the learnings of each car (inferences) to train and adjust regional models as well as to improve global general models, so that we achieve the ability to use sold cars as collectors of regional knowledge that can be applied also in scenarios that are not covered by global solutions trained in regions like the Silicon Valley.
Therefore, we require a network that enables trust and security to share critical functions as models for autonomous driving are, possibly affecting health conditions. In general, Blockchains enable the trusted sharing of information in a network to enable fast verifications of data or models to prevent data manipulations or the input of false data, which could e.g. diminish the functionality in autonomous driving. However, classical Blockchain solutions, such as Ethereum or EOS, are both extremely imbalanced towards oligarchic control of only very few strong nodes (either through computing power or (financial) stakes) and also energy inefficient, such that they cannot run inside the actual device or machine. Thus, for our distributed intelligent systems, we actually require a dynamic, adaptable and energy-efficient infrastructure for organizations and any kind of device to purposefully work together in an equally-distributed, privacy-preserving and secure network, which is why we built the eXpandable AI Network.
Learn more about our eXpandable AI Network and the Partial Proof of Kernel Work in our following article!
Besides our technical blog section, we also have a rather business-oriented section, where we introduce various use-cases of blockchain and AI technology in the industry.
More information about our business can be found on our website: www.xain.io