What is artificial intelligence (AI) and where do we encounter it in everyday life?
Depending on their research area, experts still provide slightly different definitions of artificial intelligence (AI), but they generally agree that it is the simulation of human intelligence processes by machines, especially computer systems. Colloquially, the term AI is used to describe machines that mimic “cognitive” functions that humans associate with other human minds. These processes include learning (the acquisition of information and rules for using the information), reasoning (using rules to reach approximate or definite conclusions), task execution, problem solving, and self-correction. Examples of AI include expert systems (e.g. the artificially-intelligent attorney ROSS that is now an expert system on legal issues and law), speech recognition (now on every smart phone, e.g. in WhatsApp, Amazon’s Alexa, or Apple’s Siri), and machine vision (this is not only the ability to capture a digital video or image, but also includes depth perception and correctly interpreting what is seen). Applications of machine vision, for example, include signature recognition, airplane navigation, or analysis of medical images.
Artificial intelligence can be categorized as either weak or strong. Weak AI, also known as “narrow AI”, is an AI system that is solely designed and trained to perform a particular task. Virtual personal assistants, such as Apple’s Siri, are a form of weak AI. Strong AI, also known as general artificial intelligence, is an AI system with generalized human cognitive abilities. When presented with an unfamiliar task, a strong AI system is able to find a solution without human intervention. Interestingly, as machines become increasingly capable, tasks considered to require “intelligence” are often removed from the definition of AI, a phenomenon known as the “AI effect”. For instance, optical character recognition is nowadays frequently excluded from things considered to be AI, because it has become a routine technology.
Computer science defines AI research as the study of “intelligent agents”. An “intelligent agent” is any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. To do so, the AI system needs to be able to “correctly interpret external data, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation” (Kaplan and Haenlein 2018). Kaplan and Haenlein (2018) classify artificial intelligence into three different types of AI systems: (1) analytical, (2) human-inspired, and (3) humanized AI. Analytical AI has only characteristics consistent with cognitive intelligence, meaning it can generate a cognitive representation of the world and use learning from past experience to inform future decisions. Human-inspired AI has elements from cognitive and emotional intelligence, meaning it can understand human emotions, in addition to cognitive elements, and consider them in its decision making. Humanized AI, the highest potential form of AI, should display characteristics of all types of human competencies (including cognitive, emotional, and social intelligence) and be self-conscious and self-aware in interactions with humans or other humanized AI systems. Considering that not all humans are striving to live a mindful life and interact at their best, we should be curious whether truly (emotionally, cognitively and socially) intelligent AI systems may or may not be superior to human beings in the future. – Illustrations by Gerd Altmann from Pixabay
What is federated machine learning and how can it increase security and privacy?
Former machine learning approaches required the availability of large datasets and analysis of these data sets on a central analysis server. These data sets were usually created by collecting huge amounts of data from users. Federated machine learning is a more flexible technique that allows training a model without directly seeing the data. This also means that the learning algorithm is applied in a distributed way, that there is no central point of attack for hackers, and that privacy-protected personal data is only stored locally, e.g. in case of FeatureCloud on the firewall-protected intranet of each hospital, and that such does not have to be transferred to any central analysis server. Thus, federated machine learning is very different compared to the way traditional machine learning takes place in centralized data centers. Since federated machine learning systems in general are quite flexible, they can be adapted to allow for locally personalized models. At the moment, the technology of federated machine learning is still quite new, and not many uses of it were reported by the industry yet.
In its practical application, federated machine learning is a collaborative form of machine learning where the training process is distributed among many users. A server has the role of coordinating everything but most of the work is not performed by a central entity anymore but by a federation of users, hence the name. Before the start of the actual training process, the server initializes the model. After the model is initialized, a certain number of users are randomly selected to improve the model. Each sampled user receives the current model from the server and uses their locally available data to compute a model update. All these updates of the initial model are sent back to the server where they are averaged, weighted by the number of training examples that the respective clients used. The server then applies this update to the model and so forth.
Besides increased privacy and security, there are further advantages: The model can automatically update itself without having to store the input data permanently. This makes federated machine learning extremely powerful because models can be trained with a huge amount of data, without the need for this huge amount of data to be transferred and stored on a centralized server or cloud. We can thus make use of a lot of data, in the case of FeatureCloud for example of important scientific and biomedical results, that otherwise could not have been collected or analyzed without violating a patient’s privacy. See illustration below.
What is blockchain technology and how will this “cryptocurrency”- technology aid FeatureCloud?
A blockchain is essentially a growing list of records. These records are also called “blocks”, and the blocks are protected and linked amongst each other using cryptography. By design, this makes a blockchain resistant to modification. Each block contains a cryptographic hash of the previous block, a timestamp, and transaction data (generally represented as a so called “Merkle tree”, see illustration below).
In other words, a blockchain is an open, distributed ledger that can record transactions between two or more parties efficiently and in a verifiable and permanent way. In the olden days, a ledger was made of paper and ink, and could only be stored and accessed in one place. Digitalisation and the internet enable us to access and use such ledgers openly. Typically, a blockchain is managed by a peer-to-peer network collectively adhering to a protocol for inter-node communication and for validating new blocks. Once recorded, the data in any given block cannot be altered retroactively without alteration of all subsequent blocks, which requires consensus of the network majority. Thus, blockchains may be considered secure by design and exemplify a distributed computing system, including decentralized consensus.
Originally, blockchain technology was invented by a person using the alias “Satoshi Nakamoto” in 2008 to serve as the public transaction ledger of the cryptocurrency bitcoin. The actual identity of Satoshi Nakamoto is, however, still unknown. In finance and banking, the use of blockchain technology for the first time solved the double-spending problem of digital currencies without the need of a trusted authority or central server. Besides cryptocurrency, blockchain technology has inspired and will inspire many other applications, mainly in fields where privacy, democratic consensus, fairness, secure transactions, secure record storage, secure record transfers, and permanent immutability are required. In the health care system for example, such encrypted blockchain ledgers provide a unique opportunity to store civil registry data, health records, prescriptions, financial transactions, insurance changes etc. for each patient in a secure and privacy-protected form. Patients would maintain full control over their personal data, while pharmacies, insurances, doctors, public agencies, or researchers would only be able to access as much information as the individual allows. Therefore, FeatureCloud will use this technology to increase privacy and security of primary medical data by design.