· Technical · 2 min read
Starting With Better Data
We are experimenting with retraining AI from the ground up, with one central goal: improving the quality of the data AI learns from.
We are experimenting with retraining AI from the ground up, with one central goal: improving the quality of the data AI learns from.
Our starting idea is that a stable, carefully structured environment, guided by clear rules and responsible choices, could help us develop better training material. We are exploring that idea through an experimental approach to AI, social media, and games.
Why This Matters to People
Imagine asking an AI system for help with something unfamiliar. Its answer misses your point, so you explain again. You are looking for understanding, and each disconnected reply adds another layer of frustration.
The outcome we want to work toward is simple: a person feels understood and receives useful help.
Our experiment approaches that ambition through the training data. We want to examine whether a more deliberate foundation can support clearer, more consistent examples for AI to learn from. Any improvement in the data would then need to be tested for its effect on the system's responses.
Where Argotu and Gamo Fit
Argotu, our social media platform with no algorithm, and Gamo, our game where the challenge is to fit items perfectly, are part of this wider exploration.
Their connection is the purpose behind the experiment: investigating whether different approaches to social interaction and play can help us improve AI data quality.
In Gamo, the challenge invites a simple sequence: look at an item, consider where it belongs, and find the right fit. That offers a useful way to describe the care we want to bring to training data. What belongs? What is missing? What needs closer examination?
Argotu brings that exploration into a social setting, where meaning depends on more than whether something appears to fit. Context, different perspectives, and the intention behind an exchange all deserve attention.
These projects give the experiment different directions to explore. Their value to AI training will need to be demonstrated, rather than assumed.
A Foundation Worth Testing
Clear rules and structure are the starting approach. We still need to examine whether they produce the improvements we are seeking. A more controlled environment should not be mistaken for a complete picture of people's lives or experiences.
People also deserve clarity about how any participation relates to AI training, with care for their privacy and choices.
The goal is more than building another platform or game. It is to investigate how better learning material could support AI that serves people more effectively.
That is the purpose connecting this work: improve the data, test what changes, and keep the people affected by those changes in view.