YYSI/ARTICLES / TERMINOLOGY
AI vs SI vs AGI vs ASI: What’s the Difference?
A plain-English comparison of AI, the federal SI label, AGI, and ASI—and why changing the label does not establish a capability breakthrough.
The four terms at a glance
| TERM | EXPANSION | WHAT IT TELLS YOU |
|---|---|---|
| AI | Artificial intelligence | A broad technology and research category. |
| SI | Super Intelligence | The terminology specified in EO 14434 for covered executive-branch usage. |
| AGI | Artificial general intelligence | Broad, general capability; exact thresholds differ between frameworks. |
| ASI | Artificial superintelligence | Capability substantially beyond human levels; definitions differ. |
AI is the broad category
AI can refer to systems used for tasks such as recognizing patterns, generating language, or supporting decisions. The term alone does not tell you whether a system is reliable, autonomous, or suitable for a particular job. Those questions require evidence about the actual system.
For historical context, Dartmouth’s account connects the field to the 1956 summer research project. That is why YYSI no longer describes the AI acronym as having an 85-year history. Read Dartmouth’s account of the field’s origins.
SI is context-dependent
Under EO 14434, section 3, SI initially covers the technologies in the referenced statutory AI definition. The replacement name does not itself change the model, its training, or its measured abilities. Outside that policy context, a writer may use “superintelligence” with a different technical meaning, so check how the term is defined.
AGI concerns breadth of capability
AGI usually refers to broad competence across cognitive tasks, rather than excellence in one narrow task. Google DeepMind’s explanation of its approach to AGI describes a human-comparable capability threshold across most cognitive tasks. Other frameworks operationalize the idea differently; there is no single test implied by the acronym alone.
ASI concerns capabilities beyond human levels
ASI refers to a more demanding capability concept than a label change. DeepMind’s “From AGI to ASI” research summary examines systems beyond human-level AGI, including comparisons with the capabilities of large human organizations. The point here is the distinction between a capability claim and a naming policy, not a forecast of when such systems will exist.
Has AGI become GSI?
EO 14434 does not prescribe GSI as the replacement for AGI. YYSI previously presented that pairing too confidently. GSI and fictional company-name replacements on this site are now confined to clearly marked entertainment examples. They should not appear as official terminology in a factual explainer.
A useful reading rule
When a headline says SI, ask whether it is describing an agency’s wording, a company’s branding, or a claim about system capabilities. Then look for the relevant evidence: an official directive, a company announcement, or an evaluation. The same word cannot substitute for all three.
Sources & verification
Primary sources checked October 20, 2026. Dates below refer to publication or event dates, not YYSI discovery dates.
- Executive Order 14434 — definition and implementation
- Dartmouth: The Founding and Future of AI
- Google DeepMind: Taking a responsible path to AGI, April 2, 2025
- Google DeepMind: From AGI to ASI, June 12, 2026
Written by YYSI Editorial. Analysis is identified as such. Send a correction to [email protected].