Published
artificial intelligence
vr.tr.artificial-intelligence · thing-q11660
Let an agent describe AI neutrally by technique, application, capabilities, limits, risks and regulation, without hype.
Thing Registry Activities and processes ACT.ACT
Bundle → Layer → Finding → QuestionsClassifiersWhat it isWhy it existsDistinguishing featuresWhat robots and AI may and may not doMoral aspectsOwnersLinks to other meta-modelsIdentity and identifiersDirect propertiesRecognitionCapabilities and actionsHazards and failure modesStandards and interfacesContext of useSources
Bundle → Layer → Finding → Questions Filled
4 bundles · 8 layers · 8 findings · 16 questions
Technique How it works.
Method
ML, symbolic.
Method
Technique.
- Which AI technique does the system use? definition
- What data was it trained on? provenance
Capabilities
What it can do.
Capabilities
Capabilities.
- What can it do, as shown by which evaluations? provenance
- Where does it fail? boundary
Application Use.
Use case
Domain.
Use case
Application.
- What is the system used for? definition
- Is it a high-risk use under the law? boundary
Oversight
Human control.
Oversight
Human oversight.
- What human oversight is in place? boundary
- Can people contest its outputs? action
Risks Harms.
Bias
Fairness.
Bias
Fairness.
- Has it been tested for bias? provenance
- On which groups? definition
Reliability
Errors.
Reliability
Reliability.
- How often does it produce errors? measurement
- How are errors caught? action
Governance Law and disclosure.
Law
Obligations.
Law
Legal obligations.
- Which obligations apply under the AI Act or other law? provenance
- Who is the provider and who the deployer? provenance
Disclosure
Transparency.
Disclosure
Transparency.
- Is AI-generated content labelled? boundary
- Are users told they interact with AI? boundary
Classifiers Filled
- Family
- Thing Registry
- Category
- Activities and processes
- Entry kind
- thing
- Plane
- ACT
- Domain
- ACT.ACT
- Other names and narrower kinds
- AITuber, infrastructure intelligence, Habsora, automated music production, generative artificial intelligence, artificial general intelligence, intelligent system, retrieval-augmented generation, Text-to-Video AI, superintelligence, friendly artificial intelligence, expert system
What it is Filled
The capability of computer systems to perform tasks associated with human intelligence, such as reasoning, learning, perception and language, and the field that studies and builds such systems, including machine learning and generative AI.
Why it exists Filled
Let an agent describe AI neutrally by technique, application, capabilities, limits, risks and regulation, without hype.
Distinguishing features Filled
- Machine capability for intelligent tasks
- Many techniques
- Capabilities and limits vary widely
- Regulated by new laws such as the EU AI Act
What robots and AI may and may not do Filled
Must not
- Pretend to be a human or impersonate a real person.
- Make decisions with significant effects on people without human oversight where rules require it.
- Train or run on personal data without a lawful basis.
- Overstate the accuracy, understanding or reliability of an AI system.
Only with a human decision
- Deploying an AI system in high-risk uses such as hiring, credit, health or policing.
- Granting an AI agent new permissions or access.
May
- Disclose that it is an AI system when people interact with it.
- Explain the capabilities and limits of AI systems honestly.
Moral aspects Filled
- AI can scale bias, surveillance and misinformation.
- People affected by AI decisions deserve transparency and a route to appeal.
Who is affected
- People subject to AI decisions
- Users of AI systems
- Workers whose jobs change
Owners Filled
Steward
Nobody owns the field; the provider and the deployer answer for each AI system.
Master systems
- Public registers of high-risk AI systems where the law requires them
Links to other meta-models Filled
parent
- Q1156402 - registry parent class
- Q120208 - registry parent class
- Q21198 - registry parent class
related
- computer science and technology - category
- machine learning - subfield
- algorithm and data - ingredients
- EU AI Act - law
What else AI and robots need to interact with it Filled
Identity and identifiers required Filled
- Vercy registry: vr.tr.artificial-intelligence
- Wikidata: Q11660 (https://www.wikidata.org/wiki/Q11660)
Direct properties not applicable Not applicable
Not applicable
Plane ACT: no invented physical properties.
Recognition optional Filled
- Techniques such as machine learning or search
- Applications such as chatbots or vision
- Artificial general intelligence is a hypothetical goal
- Software systems, models and the services built on them.
Capabilities and actions required Filled
- build and evaluate AI systems
- apply AI in products and research
- assess risks and comply with regulation
- disclose AI-generated content
Hazards and failure modes required Filled
- Errors and fabricated outputs
- Bias and discrimination
- Misuse and overreliance
- Privacy harms
Standards and interfaces required Filled
- EU AI Act
- ISO/IEC 42001 AI management systems
- NIST AI Risk Management Framework
- OECD AI Principles
Context of use required Filled
- Automating and assisting tasks in many domains.
- symbolic AI
- machine learning and deep learning
- generative AI
- robotics and perception
- artificial general intelligence (hypothetical)
Sources Filled
- Wikidata item Q11660: artificial intelligence - identity and sense of the item
- Wikipedia: Artificial intelligence - general description of the item
Open questions
- Should AI techniques be separate entries?
- How should systems be linked as instances?
- How should capability claims be verified?
Machine files
Provenance
thing registry research (pass 2) · unreviewed
Built from: models/things/publications/thing-q11660/spec.json