fleetingswallow.comarticleAchieving AI-style reasoning without an LLM
Minotaur is a GPUless reasoning architecture built on ELC (Energy, Control, Logic) typed roles and the MINOS representation. One "settling" operation over one network handles deduction, abduction, diagnosis, seeing versus doing, belief revision and order-dependent histories. The authors report it matched a language model on a blind police-interview test (30 of 30) and on learned production-line faults (10 of 10), and that it recovered Aristotle's 24 valid syllogisms.
TL;DR Describes Minotaur, a non-LLM reasoner that runs deduction, diagnosis and causal queries as one network-settling operation on a single CPU core, with self-reported blind-test results.
Organisations named here
- Sapientronic
- Digital Equipment Corporation
- Stanford University
- British government
- Fifth Generation Computer Systems project
- Douglas Lenat
- ELC
- Fifth Generation Computer Systems
- General Problem Solver
- Judea Pearl
Technologies named here
- Minotaur
- ELC adapter
- ELC neural network
- Minotaur exact side
- Minotaur ELC side
- Prolog
- SAT solvers
- Minos
- LLM
- Cyc
- Mycin
- XCON
- Logic Theorist
- Bayesian network
- ELC
- SAT solver
Techniques named here
- Settling
- Crossings
- Seeing versus doing
- Chain of worlds
- Fault supposition and diagnosis
- Learning typed roles from data
- Transport
- Next-word prediction with step-by-step reasoning
Related in research
