Epistemic & Machine Learning Diagnostic

AI Pattern vs Intelligence Explorer

Inspect why statistical next-token prediction mimics conscious reasoning. Test prompt counterfactuals, explore token probability trees, and isolate where statistical co-occurrence diverges from physical and causal reality.

Prompt & Grounding Probe

Causality Trap
0.70
Next-Token Probability Distribution & Branching Path
P(w_{t} | w_{<t})
Syntactic Fluency
98.4%
Grammar & style adherence
Grounding Gap Index
74.2%
Plausibility vs physical reality
Entropy (Η)
1.84 nats
Branching dispersion
Statistical Co-occurrence Hallucination Detected
The model selected high-probability n-gram associates ("purred and fell asleep") because "cat" and "mat" co-occur millions of times in training corpora, completely missing the thermodynamic invariant that liquid magma vaporizes organic matter instantly.
Historical & Cognitive Dimensions

Dartmouth 1956 Naming Shift

McCarthy Branding Matrix

In 1955, John McCarthy coined "Artificial Intelligence" to secure workshop funding at Dartmouth. Compare how human psychological projection shifts based on framing:

If Named "Applied Statistical Prediction"

Public perception: Autocomplete, weighted matrix operations, high-dimensional curve fitting.

Projection Index: Low (12%)
Branded "Artificial Intelligence"

Public perception: Machine minds, thinking systems, sentience, synthetic sapience.

Projection Index: Severe (89%)

Humans evolved a hypersensitive agent-detection bias: when text looks grammatically coherent, the brain reflexively assumes a sentient author behind it.

Cognitive Capability Taxonomy

Disparate Functions

Click a quadrant to inspect its grounding substrate, semantic awareness, and failure boundaries:

LLMs / Autocomplete
Pattern Co-occurrence
Statistical text continuation. High fluency, zero physical causality grounding.
Expert Systems
CT / Anomaly Detection
High domain-specific accuracy, non-generative, no self-model or conversational persona.
Symbolic Solvers
Formal Logic & Math
Guaranteed axiomatic consistency (e.g. Lean, Z3). Incapable of fluid linguistic synthesis.
Biological Sentience
Embodied Sapience
Homeostatic survival drive, somatic grounding, true experiential comprehension.
Currently Inspecting: Large Language Models (LLMs)
Mechanism: Calculates softmax over billions of parameters. Knows that words fit together syntactically without experiencing the referent objects.

Next-Token Statistics vs Physical Grounding

Read the explanation

Large language models evaluate sequential token logits across massive training corpora, calculating softmax probabilities over statistical co-occurrences. In the liquid magma prompt, high co-occurrence between cat and mat favors purring, completely ignoring thermodynamic reality. Switching the evaluation paradigm to Physical Grounding enforces thermodynamic constraints, elevating physically valid outcomes and closing the epistemic gap.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.