We are building Symbolic Superintelligence.
Linguistic Agents develops a unique approach to artificial intelligence that we call Symbolic Superintelligence.
The tripartite Faculty of Language
Linguists naturally see Syntax as the starting point: C-I ← Narrow Syntax → A-P.
But in AI, the direction of generation is:
And the direction of comprehension is the other way around:
The three engines implement Narrow Syntax
In the Linguistic Agents architecture, Narrow Syntax is implemented by three syntactic engines: Inner Syntax, A-bar Syntax, and Distributed Morphology (DM).
Together with C-I and A-P, these form five engines working in two directions: generation and comprehension.
The full path from agent to agent
A full communication path goes from one agent to another through ten engine uses: five in generation, and five in comprehension.
The bracketed names are formats. The almost-white names are engines. speech/text is the external format that can be sent.
This is the central communication path in Symbolic SI. The agent does not communicate by sending Agentic Form directly. Agentic Form is encoded through the engines into speech/text, and another agent uses the corresponding engines to recover Agentic Form on the receiving side.
Symbolic SI is not just a grammar pipeline
In Symbolic SI, agents use Agentic Form in all their work with other Cognitive Faculties.
The Five Engines handle the Faculty of Language. In comprehension, they deliver Agentic Form to the agent. In generation, the agent gives Agentic Form back to the Faculty of Language, where it is encoded as speech/text.
Some of these areas are also studied by linguists, especially where language, discourse, assumptions, and interpretation meet.
- discourse management
- shared assumptions
- background knowledge
- presuppositions
- implicatures
- search
- memory
- planning
Some of these areas are also studied by linguists, especially where language, discourse, assumptions, and interpretation meet.