WILLIAM

Responsible: Ben Goertzel

Papers: Hyperon for AGI⇒ASI Whitepaper (2025), §5.12

Status: Experimental. The whitepaper describes WILLIAM as being integrated into MORK infrastructure. The core compression-based pattern detection is under development; the full integration with PLN scheduling, ECAN attention, and neural network acceleration described below remains a research goal.

WILLIAM is an adaptive-compression-based approach to cognitive pattern discovery. It embodies the principle that the patterns worth remembering are those that compress experience most effectively.

Core Idea

WILLIAM acts as a cognitive feature detector that continuously asks "what's the simplest explanation that captures what I'm seeing?" The whitepaper describes moving WILLIAM directly into MORK's trie infrastructure, aiming to transform it from a batch analysis tool into a real-time guide for reasoning and neural processing.

Proposed Technical Implementation

The whitepaper describes adding instrumentation to MORK's trie nodes: local occurrence counts, subtree totals, compression-gain sums, and ranked lists of top-k children. This would enable weighted iterators returning heavy subpatterns directly from any point in the graph. The proposed API includes iter_prefix_topk(prefix, k) and iter_any_topk(k) operations.

Intended Consumers

The whitepaper describes these iterators serving PLN (prioritizing high-value subgraphs), backward chaining (following "heavy edges"), schedulers (compression-adjusted priorities), and ECAN (compression-driven importance signals).

Neural Network Acceleration (Proposed)

The whitepaper describes potential integration with transformers and predictive coding networks, where WILLIAM could find "heavy-hitter features" to prioritize computation, trigger on-demand refinement, and prune low-value frequency bands.

Technical Deep Dive: WILLIAM Full — MORK trie instrumentation, iter_prefix_topk/iter_any_topk API, compression-gain metrics, wait-free concurrency, and neural heavy-hitter acceleration.

Key References

• Goertzel, B. (2025). Hyperon for AGI⇒ASI Whitepaper, §5.12: WILLIAM-on-MORK
• WILLIAM is also discussed as a cross-stack mechanism: Cross-Stack Research Directions