HALF ECHO / CLUARTIFICIAL INTELLIGENCE / EXPLAINED

EXPLAINED IN PLAIN ENGLISH / TOPIC 02 / 08

How does AI work?

Many AI systems are trained using examples. Software adjusts a mathematical model to identify useful patterns, then applies those patterns when processing new information.

CLU / EXPLORE & UNDERSTANDILLUSTRATIVE / NOT A LIVE AI SERVICE

Follow an example through the system.

TRAINING / PREPARE

Examples

Labelled animal picturesTraining and evaluation data

For a supervised image classifier, people assemble and label example images. Separate evaluation data helps test performance beyond the examples used to fit the model.

Illustrative supervised-learning workflow. AI also includes other learning methods and rule-based systems. No model is running here.

01

Training versus using AI

Training is the process of adjusting a model from examples. Inference is the name for using a trained model to make a prediction or produce an answer. They are different activities.

02

A simple example

Imagine showing a system thousands of labelled pictures of different animals. During training it learns visual patterns associated with those labels. Later it may estimate which animal is in a picture it has not seen before.

03

What about chatbots?

Language models work with pieces of text called tokens. They generate responses using patterns learned during training and the information available in their context. Their answers may sound confident even when they are inaccurate.

CHECK THE ORIGINAL MATERIAL

Sources & further reading

Starting points for the explanation and the illustrative interface above. Sources are external; examples are not CLU capabilities.

← ALL CLU TOPICS

THE PROJECT BEHIND THE RESOURCE

CLU / Codified Likeness Utility

Neil Hamson’s personal Artificial Intelligence research and software project under Half Echo. This educational resource shares a separate aim: helping people understand AI through accessible explanations and credible evidence.

RESEARCH DEVELOPMENT / CEASED PENDING FUNDING & RESOURCES

No authorised restart date, public AI service or commercial subscription. CLU has not trained a foundation language model.

INSPECT THE PUBLIC RECORD

Evidence before promises.

The public repository contains curated documentation. Private application code is not published. Source presence, reported verification and demonstrated operation are distinct evidence states.

EXPLORE CLU PUBLIC EVIDENCE ↗CONDITIONAL DEVELOPMENT ROADMAP ↗

Future work depends on sufficient funding and technical resources. No funding target or investment terms have been approved.

DISCUSS COLLABORATION OR POTENTIAL SUPPORT ↗