Our definitions for various terms are short and meant to be consulted quickly. For a fuller understanding of a term, at the end of our definition, we include either a link to the term's Wikipedia entry or a link directly to the source of the term. For a more general and more comprehensive view of AI and the terminology used in AI, reader may consult Artificial Intelligence, Dreams and Fears of a Blue Dot. That website also has a larger Glossary of Terms, all AI related.
We use this term in a milder way than usual, namely that governmental form where humans are still in control of our institutions, but they do it with extensive use of AI. Loosely, we refer to it as "Government on a Chip". In a stronger way, not ours, the term refers to AI being in complete control. [Wikipedia]
The intelligence present in the machines (hardware or software) we produce. [Wikipedia]
All throughout SD-AI you will encounter the terms Machine Learning (ML) and Deep Learning (DL), the two most important subfields of AI. The following diagram shows the relationship of AI with these two subfields.
The second set of terms you will encounter often is shown in the diagram below. These terms are also in the Glossary below, following immediately the diagram.
The AI functioning at the same level with human intelligence, capable of solving all tasks that humans do. The relationship with the other types of AI. [Wikipedia]
Also known as Weak AI, this is AI that is implemented for one narrow task; it can achieve a higher level of intelligence than humans, on that narrow task. [Wikipedia]
The AI which, after having surpassed the level of human intelligence, functions at such high level that it is unimaginable to humans. [Wikipedia]
This term is not standard, we simply need it in order to anchor and make more precise most of the powerful AI systems on which we will base our work. AWI systems work on large graphs, are based in large data centers, and accomplish many ANI tasks. They only have cognitive powers, and while they may understand human emotions and human consciousness, they will not exhibit those properties themselves. AWI shows a different possible progression from ANI to ASI, without having to go through AGI.
Mixed reality; it combines physical reality with the Virtual Reality the user experiences within a computer program. [Wikipedia]
A linked list of chronologically ordered records, with each link based on cryptography, and the technology used to produce, verify and distribute these lists. [Wikipedia]
Due to its decentralized core characteristic, i.e. the intentionally designed lack of a central authority, blockchain cannot be governed with by the usual frameworks, which are all based on a central source of authority. There have been many attempts to provide such a framework, and they all have merit. The reader is better off googling for a variety of such sources. Here is one for example.
BLOOM is a Large Language Model (LLM), trained to continue text from a prompt on vast amounts of text data using industrial-scale computational resources. As such, it is able to output coherent text in 46 languages and 13 programming languages that is hardly distinguishable from text written by humans. BLOOM is open-source and partly because of this, it will be our reference LLM. Worth getting familiar with it:
The LLM most talked about these days. It is being a bit sensationalized both in the mainstream media and on social media, and although there is still much work to be done, it has achieved already a level of conversational prowess unmatched by any other AI system that we know of. [Wikipedia]
Authority and decision making being distributed among peers, as opposed to a central point. In our work we usually refer to a more specific instance of decentralization, namely blockchain decentralization. [Wikipedia]
The participants (=members) of a DAO do not report to a traditional hierarchical structure. The DAO is being run according to rules that are agreed on by the DAO members on a peer basis and recorded on a blockchain.[Wikipedia]
A subfield of Machine Learning focusing on artificial neural networks having more than one hidden layer. The AI Components diagram shows the relationship of AI with Machine Learning and Deep Learning, the two most talked about subfields of AI. [Wikipedia]
Same as E-democracy, described below.
Citizens of the country decide on legislation proposals directly, not through elected representatives. Switzerland is the best known example. We dedicate a page to how AI and democracy interact in Switzerland.[Wikipedia]
A very important term for our mission, close to what we are trying to do. It means a broader participation of the population in the political and governing process through the use of technology. We have a more limited view of this participation, namely through the development of practical AI-based applications leading to stronger democratic institutions.[Wikipedia]
This is AI whose models can be understood by humans (explained to humans). But most of the spectacular success of AI nowadays has been with "black-box" AI where explainability plays no role and results are left to speak for themselves; that is especially the case with neural networks. There is however a strong movement nowadays away from the "black-box" approach and towards a much more transparent AI, sometimes known as interpretable AI. All three approaches (black-box, explainable, and interpretable) have their adherents. [Wikipedia]
"Follow is a decentralized social protocol based on blockchain technology. By providing complete basic development tools and a standard smart contract interface, any individual or organization can build applications based on the protocol. In any application built on Follow, users not only have full control over their own social identity and data but can also interconnect and combine various applications, such as building decentralized finance Lego to form a decentralized, hybrid and hierarchical SocialFi network with an autonomous economic system". The quote is from the press release:[Follow, the first decentralized social protocol designed for Web 3.0, is live]
For us this a practical implementation of the "programs as proofs" equivalence, i.e. the development of software applications using tools of mathematics. See our library entry describing it in more detail. [Wikipedia]
For our purposes, the term will be used interchangeably with the term "AI Revolution". But in general, the term has a broader definition, including other technologies like gene editing. [Wikipedia]
The European regulation for data protection and consumer privacy, it has become the golden standard for many other countries. In the US, there is no federal regulatory equivalent; at state level, the California Consumer Privacy Act (CCPA) has been drawn for similar purposes. [Wikipedia]
Their front page: "We're on a journey to advance and democratize artificial intelligence through open source and open science. The AI community building the future. Build, train and deploy state of the art models powered by the reference open source in machine learning." [Wikipedia]
Read this for a good description.
This is AI used for high stakes decisions where results have to be justified by humans for humans. Most of the spectacular success of AI nowadays has been with "black-box" AI, especially with deep learning. Then we had explainable AI, whose models can at least be explained by humans to humans. Interpretable AI is an even more transparent form of AI, one whose models can not only be explained but built from the beginning with constraints that ensure that results can be understood by humans. [Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead]
The subfield of AI studying the statistics-based algorithms which develop models they learn from data. The AI Components diagram shows the relationship of AI with Machine Learning and Deep Learning, the two most talked about subfields of AI. [Wikipedia]
The term has been hyped up and therefore has generated a suspicion that it is nothing but a buzzword. But this is an over-reaction. The term has merit and it will become a reality in stages. Think of it as an extension of what happens with players in an online game. Players are represented by their avatars which they control. There isn't much of a stretch to extend that situation to social media and even to work environments. [Wikipedia]
NLP is a multidisciplinary field, but we are only interested in the AI algorithms that are used for various NLP tasks, like language translation, chat bots, personal assistants, etc. In particular, neural networks and word embeddings based on these neural networks have completely transformed the field and account for most of its current spectacular successes. [Wikipedia]
A fairly small AI research company having a large impact on the direction of AI. Developed some of the most powerful AI models at this time, ChatGPT and DALL-E 2 among them. And it allows access to them through an API. It was started as a nonprofit organization, but it now has a for-profit component.[Wikipedia]
A decentralized software development model that promotes the distribution of various software applications and their documentation freely to the public. Open-source projects are also produced in a transparent and collaborative way by teams that may be geographically very dispersed. Because of its nature, open-source tends to be of higher quality than proprietary source code. [Wikipedia]
META made a very significant decision, namely to allow the entire AI community — academic researchers, civil society, policymakers, and industry — access to OPT-175B, their latest LLM. The stated goal is to "develop clear guidelines around responsible AI in general and responsible large language models in particular, given their centrality in many downstream language applications. A much broader segment of the AI community needs access to these models in order to conduct reproducible research and collectively drive the field forward. With the release of OPT-175B and smaller-scale baselines, we hope to increase the diversity of voices defining the ethical considerations of such technologies". [META's Announcement]
Citizens of the country do not decide directly on legislation proposals. These decisions are made through their elected representatives. Almost all modern democracies are representative democracies. In SD-AI we mostly use the US constitutional republic as a model for representative democracy. [Wikipedia]
An AI that would show an ability to experience sensations and feelings. Despite the more sensational claims made today, such an AI does not exist yet, and it is an open question if it will ever. For sure it will learn to recognize that humans have feelings and sensations. But having such inner experiences itself is a different matter. [Wikipedia]
A computer-generated simulation of a virtual, interactive, immersive, three-dimensional environment with which users interact through specifically designed equipment [Wikipedia]
A somewhat controversial term and a bit hyped up at the moment (Dec 2022). Some take it to be a decentralized blockchain-based web with all sorts of smart contracts flying around ([Wikipedia]), others take it to be the Semantic Web ([Wikipedia]). For us it will be both, because it is likely that AI will fuse these two definitions into one.