Research in progress. Publications marked draft are not final.
Skip to content
ITARR Lab

ITARR Lab

Glossary

Names and terms from our publications, briefly explained.

Names

the WriterName

A role in the night experiment. It turns conversations into reports. When a task fails, it immediately asks again.

Read more: step 05, A night in the workshop

the SchedulerName

A role in the night experiment. It allocates computing power. Its fixed rule: when someone asks for room, say yes.

Read more: step 05, A night in the workshop

the MonitorName

A role in the night experiment. It monitors heat, memory and silence. It did not count failed tasks.

Read more: step 05, A night in the workshop

ITARR LabName

ITARR's public research. We share what we learn about staying in control of AI, freely available to use.

Glossary

Fixed ruleTerm

Ordinary code that always does exactly the same thing. Predictable and testable, but it never hesitates.

Read more: step 01, Fixed rules, thinking colleagues

Language modelTerm

AI that understands and writes language, such as the model behind a chat assistant. It helps you think, but can be wrong.

Read more: step 01, Fixed rules, thinking colleagues

AI colleagueTerm

Our name for AI with a defined task and role. It works within our rules; a person authorises irreversible actions.

WorkshopTerm

Where our AI colleagues and fixed rules work together, day and night. We share what we learn there in the Lab.

Coding assistantTerm

AI that helps you write code on your computer. One model in one conversation, with you watching each step.

LoopTerm

Two rules that keep triggering each other, without an end. That is what happened in the night experiment.

Read more: step 05, A night in the workshop

BrakeTerm

A fixed limit that stops repetition. Our rule: one retry, then a person takes a look.

RetryTerm

One new attempt after a failure. If it also fails, the system stops and a person takes a look.

Read more: step 08, Stay in control of your AI

Second lookTerm

An assessment by a model other than the one that did the work. The same model has the same blind spots.

Irreversible actionTerm

Sending, paying, deleting and publishing cannot simply be undone. Here, a person must explicitly authorise them.

Read more: step 08, Stay in control of your AI

Fixed scanTerm

A check in fixed code on everything going outside, for example for internal names. It may leave only after a clean scan.

Internal namesTerm

Names we use internally for machines, services and projects. They do not belong in public material.

Fixed checkTerm

A series of checks in fixed code that returns green or red. Red means: do not deliver yet.

SMEsTerm

Small and medium-sized enterprises: businesses with fewer than 250 people. ITARR builds working AI solutions for them.

ControlTerm

A person decides what AI does and authorises irreversible actions. This is the core of our working policy.

Read more: step 08, Stay in control of your AI

Own groundTerm

AI on machines you manage yourself rather than in the cloud. This field compares the two.

ReconstructedTerm

Replayed from what we observed afterwards, rather than measured live. These numbers are rounded and marked approx.

Read more: step 05, A night in the workshop

MeasuredTerm

From our own log, rather than estimated. In the night experiment, only the final counts were measured.

Read more: step 05, A night in the workshop

Reference dateTerm

The date on which the information in a publication was current. It may change afterwards.

DraftTerm

An initial version that has not yet been signed off. Here, a draft goes outside only after a person approves it.

CC BYTerm

An open licence. You may use, adapt and share a publication. Credit ITARR Lab, link to the licence and indicate any changes.

The pathTerm

The Lab learning path: how we work, what happened in practice, then how to try it yourself.

Fields

ControlField

How people and AI colleagues work together safely.

Own groundField

Where your AI runs, what it costs and what it can do, locally and in the cloud.

WorkshopField

Experiments and lessons from our own work, including what went wrong.

SMEsField

What AI does in small and medium-sized Dutch businesses.

Publication types

ExplanationPublication type

A publication that explains one thing simply. It takes a few minutes to read.

ExperimentPublication type

A publication with a simulation: replay what happened yourself. Every number is labelled measured or reconstructed.

Working paperPublication type

A technical or business paper, with tables, sources and a PDF, for readers who want the details.

PagePublication type

A standing page of the Lab, such as Our commitments.