Toward the Unknown

What is curiosity, really?

The inclination to travel in the direction of the unknown rather than away from it is known as curiosity. Anything that is fresh gets poked, tasted, opened, and disassembled by humans, particularly children. Many animals follow suit: crows examine shiny things, dogs sniff strange objects, and cats look into every box that comes into their domain.

Curiosity is fundamentally about investigating things that don't yet meet our expectations. Although it doesn't always have a clear purpose right away, it eventually aids in an organism's learning and adaptation to its surroundings.

For this assignment, the question is if artificial creatures exhibit curiosity or at least a plausible simulation of it.

Real robots that behave curiously

1. The Mars Curiosity Rover

NASA's Curiosity rover on Mars is a very real example. The exploration of the unknown is the foundation of its entire goal. The rover is outfitted with cameras and scientific sensors that continuously search the surroundings for intriguing rock formations, peculiar soil colors, or chemical clues that could indicate the presence of water or organic material.

The rover may alter its course, approach, and take further measurements when it notices something unusual. From our perspective, that appears to be a creature that "noticed something weird" and approached to investigate. Naturally, it is according to instructions and onboard software, but the behavior is still quite similar to animal curiosity-driven exploration.

2. Curiosity-driven learning agents

Reinforcement learning agents in machine learning are specifically rewarded for discovering novel states. Intrinsic curiosity modules, in which the agent receives an internal incentive anytime it encounters something unexpected or enters an environment it has never encountered before, have been tested by OpenAI and others.

These agents actively seek out new aspects of their surroundings rather than simply repeating the same behavior that clearly yields an external reward. They explore uncharted territory in virtual mazes, attempt strange game movements, and occasionally uncover tactics that were never explicitly intended. Here, curiosity is a mathematical signal rather than an emotion, yet the behavior that results is once more strikingly comparable to how live things investigate.

3. Self-mapping household robots

A more common example would be a Roomba-like robot vacuum with room mapping capabilities. It doesn't take a set route when it initially reaches a new area. It travels, collides with objects, makes turns, and progressively creates a map of barriers and open spaces. When furniture is moved, it will detect that anything doesn't match its internal map and update it by exploring once again.

This resembles a little animal cautiously exploring a new area from the outside. Although it doesn't "wonder" about it like humans do, the robot is responding to novelty and uncertainty in its environment.

Why this feels like curiosity

In all three examples, artificial creatures:

notice when something is new or unexpected,
move toward that thing instead of ignoring it,
and update their internal model of the world afterwards.

That pattern closely resembles how curious people and animals act. The distinction is that we experience emotions and a subjective sense of amazement when we are curious. It is implemented as sensor readings, prediction errors, and incentive functions for robots and agents.

However, observing these systems makes it difficult to avoid seeing some sort of curiosity in them, or at the very least, a convincing artificial replica of it.

A curiosity-driven creature

This artificial creature is a tiny robot equipped with a novelty detector and a simple proximity sensor. The robot continuously scans its environment, comparing its senses to a very basic internal memory of what it has previously seen.

One rule governs its behavior: Go in the direction of the thing it is least familiar with.

The robot turns and moves toward an object, sound, or location it has never seen before. The robot stores the new information in memory and moves on to the next unfamiliar object once it is close enough to register it correctly.

The creature's only objective is to lessen uncertainty in its surroundings. It simply approaches novelty and updates its tiny memory without picking up objects or thoroughly analyzing them. The end product is a very basic form of curiosity: a robot that always advances in the direction of its least familiar object.

Curiosity, even in its most basic form, causes a discernible change in behavior: move toward what you don't yet know, update your knowledge, and keep exploring. Although the movement appears remarkably similar, the robots in these examples follow this pattern through code and sensors, whereas humans and animals follow it out of instinct and curiosity.

Curiosity seems less magical and more universal when it arises from such basic causes. It implies that exploration is a highly effective method of learning rather than merely an emotion. And if both machines and living creatures rely on it, then curiosity might be one of the most fundamental ways any creature begins to make sense of the world.

Page updated

Google Sites

Report abuse