Research preparation
Explore terminology and competing explanations, then verify the actual literature and identify unanswered questions.
Graduate students & researchers
Develop, test and reproduce research.
American Super Intelligence will give graduate programs, public laboratories and smaller institutions more routes into model research. Teams will use public models to test ideas, write scientific software and reproduce experiments.
A proposed American Super Intelligence use case.
For research leaders and builders
Practical uses
Reproducible analysis
Draft analysis code, write tests and document workflows using permitted data and methods.
Model science
Study Liberty OSS models, reproduce baselines, evaluate adaptations and propose controlled model-training experiments.
AN EXAMPLE IN PRACTICE
A research group wants to know whether a small model can classify a particular kind of public document.
- The team defines a task, rights-cleared dataset and held-out evaluation before changing the model.
- They compare an unchanged baseline with an approved adaptation and record data, configuration, resources and failures.
- Another researcher reproduces the result and documents where the improvement does not hold.
THE HUMAN CAPABILITY
What people will learn
- Establish provenance, licensing and reproducible methods.
- Distinguish useful results from benchmark artifacts.
- Account for failed runs, uncertainty and full resource use.
THE DELIVERY PATH
How access will work
LibertyGPT will support general research preparation. Liberty OSS releases will support independent evaluation and adaptation. Research compute will have a separate allocation process, with the team, scope and resources matched to the experiment.
Understand the model familiesHow success will be measured
Look for reproducibility, scientific value, participation beyond well-funded institutions and credible negative results as well as improvements.
Requirements for this use
No unlimited compute allocation or frontier research result is promised. Restricted government and Liberty Sentinel models do not become public research resources by default.
This scenario describes the proposed experience. Product evaluation and the responsible organization’s approval will precede deployment.