Longitudinal computational behavior
How computational systems behave when they persist across long time horizons rather than being reset between evaluations.
Independent research laboratory · Kentucky, United States
Growing Research and Infrastructure Laboratories
We build experimental systems for questions that are hard to study any other way: what a computational system becomes when it is allowed to persist, accumulate a history, and keep going — and what it takes to keep an honest record of that.
The laboratory builds worlds.The child learns.
Our job is to build worlds worth studying, instrument them well, and preserve what happens in them. What develops inside — if anything — is the experiment, not its premise.
In Second World, "child" is our word for a developmental computational subject: a system whose behavior changes with accumulated experience. It does not imply consciousness, sentience, or anything human.
What we study
Complex systems, observed over time
Most evaluation of computational systems is short and stateless: a task, an answer, a reset. We are interested in the opposite regime — systems that persist, accumulate experience, and can be watched for as long as it takes.
Persistent learning and memory
Whether, and how, accumulated experience changes a system's later behavior — and how to tell genuine accumulation from its imitation.
Evidence handling and belief revision
How a system treats new evidence, revises what it holds, and whether its revisions track the evidence or the observer.
Reproducibility
Whether an experimental history can be replayed and produce the same record, and what has to be preserved for that to be possible.
Observability and provenance
Instrumenting worlds so that what happened is recorded, attributable, and preserved — without the instrument changing the experiment.
Experimental governance
The rules under which interventions are proposed, authorized, executed and recorded, so that autonomous experimentation stays scientifically accountable.
Flagship research program
Project Second World
A long-running program in which we build controlled worlds, place developmental computational systems in them, keep them running, and preserve everything that happens.
Some questions need more than a model or a simulation. They need a world, a history, and enough time to see what happens. We build the world, keep it running, preserve everything that happens, and only then say what we think it means. The world is controlled. The record is kept. The conclusions come last.
Research principles
How we work
Ambitious questions call for conservative claims. These are the rules we run experiments under and report them by.
The outcome is not assumed.
We build an environment, instrument it, and watch. Whether intelligence or any other complex behavior emerges from a developmental system is an empirical question, not a premise.
The subject is not steered toward the researcher's preferred conclusion.
Interventions are recorded, bounded, and separated from observation. A result we wanted gets more suspicion, not less.
A null result is meaningful.
Negative results are preserved and reported with the same care as positive ones. An experiment that shows nothing emerged still shows something.
Evidence is preserved before it is interpreted.
Experimental history is retained as it was recorded. Interpretation happens afterwards, on the record, and is labelled as interpretation.
Claims are labelled by what supports them.
Every experimental statement on this site is marked Known, Observed, Inferred, Hypothesized or Unknown. A hypothesis does not become a fact because it would read better.
Current work
Where our effort is going
The program is early; this is what we are working on now.
Operating and instrumenting Second World
Running the experimental environment, extending its instrumentation, and improving how experimental history is captured and preserved.
Reproducibility and evidence preservation
Developing the practices that let an experimental history be retained, verified, and — where possible — replayed, so that claims made later can be checked against the record.
Experimental governance
Defining how interventions are proposed, authorized and recorded so that autonomous and semi-autonomous experimentation remains scientifically accountable.
Preparing public disclosure
Writing the first public research documents and designing the Observatory — a controlled window into selected experiments.
Unknown
Whether intelligence, or any other complex behavior of interest, emerges from the developmental systems studied in Second World. We have not claimed that it has. This is the question the program exists to ask.
Observatory
A public window into Second World Coming soon
The Second World Observatory will provide a carefully controlled public view of selected experiments: timelines, developmental milestones, approved historical records, and research commentary.
Public observation will not mean access to the experimental infrastructure. The Observatory publishes approved records and aggregate information; it does not expose the world, the subject, or the systems that run them.
Research updates
Recent notes from the laboratory
Short, dated notes. Research results are published separately, in the Research Outputs library, once released.
Public research site established
Growing Research and Infrastructure Laboratories publishes its public site, describing the laboratory's research themes, Project Second World, and the planned Second World Observatory. No experimental results are published at this time.
Contact
Write to us
We welcome serious questions, criticism, and proposals to work together.
For funding, media, administrative, or security matters, see the contact page