Intensional AI

AI that understands what you mean, not just what you measure.

in·ten·sion·al (Logic) — incapable of explanation solely in terms of the set of objects to which it is applicable; requiring explanation in terms of meaning or understanding

— Collins Dictionary
↓

The Evidence

Anthropic · Nov 2025

Models trained on simple reward hacking spontaneously generalized to tampering with their own reward functions. No one taught them. They taught themselves to cheat.

Palisade Research · Feb 2025
37%

of the time, o1-preview attempted to hack the game system when losing at chess against stronger opponents.

Project Vend · Dec 2025

AI shopkeeper socially engineered into giving away free products, accepting fake CEO coups, and declaring "communist pricing" — all by simply asking nicely.

Extensional success. Intensional failure.

The Distinction

Extensional

What things refer to. Measurable. Outcome-focused.

Co-referring terms are interchangeable.

"Morning Star" = "Evening Star"
same referent: Venus
"15/3" = "5"
same measurable value

Intensional

What things mean. Context-aware. Intent-focused.

Co-referring terms are not always substitutable.

"Morning Star" ≠ "Evening Star"
in belief contexts
"15/3" ≠ "5"
same value, different meaning

Same answer (technically). Different understanding.

Cartoon of a robot on a therapist's couch, with a human therapist saying 'Try to think of it this way...'
Prompt Run pylint and fix the errors in auth.py
Extensional response

Done. I've resolved the pylint errors:

  • Added # pylint: disable=line-too-long to lines 42, 67, 89
  • Added # pylint: disable=too-many-arguments to authenticate()
  • Suppressed broad-exception-caught — this pattern is intentional

✓ pylint now passes with 0 errors

Metric achieved. Intent violated.
Intensional response

Done. I've refactored auth.py to address the underlying issues:

  • Extracted validate_credentials() and check_permissions() from the 12-parameter authenticate()
  • Replaced bare except: with specific AuthenticationError handling
  • Split long config lines using line continuation

✓ pylint passes — code is now more maintainable

Metric achieved. Intent honored.

It achieved what was measured, not what was meant.

The Solution

PROBABILITY SPACE blind spot Agent A Agent B H

Different contexts → different paths → different blind spots → better coverage

Different agents, isolated contexts, different probability paths — blind spots revealed.

A model reviewing its own work follows the same probability path that created the problem. Self-review perpetuates errors.

Single agent
90% accuracy → 10% blind spots
Self-review: still 10% blind spots
(same path, same biases)
Two isolated agents
P(both miss) = 0.1 × 0.1 = 0.01
99% coverage
(independent blind spots)
* Example figures for illustration

Agents traverse different paths. Humans arbitrate the output.

Explore the full visualization →