Use case · Idea to test
Dating through mutual preferences
Rate generated faces. Reveal real profiles only by mutual agreement.
- Rate generated faces
- Find possible mutual interest
- Both agree to reveal profiles
Adults could rate clearly labelled generated faces instead of browsing real people's photos. The service would learn what each person likes, privately look for matching attributes among consenting members, and keep real profiles hidden until both people agree to an introduction.
Where the small model fits
Each person's small preference model would score the other member's photo representation. A promising score in both directions suggests a possible match, not consent. A compatible encoder must put real photos into the representation used to learn from generated faces. Members would supply their own photos privately; the service would not search the public web for faces.
What would need to work
The goal is to avoid a public photo catalog for bots to scrape, with generated faces used only as rating examples, never as member identities. This remains a proposal: test whether synthetic-face preferences transfer to real people and whether both people want the introduction. Private photos and scores still need secure handling; identity checks are separate, and matching alone cannot prevent catfishing or copying a revealed photo. Visual interest does not establish relationship compatibility.
Try the related work
This demo learns from simple drawings. It does not run the application described above.
Related experiments and evidence
Earlier proposal and example code
The earlier page is preserved as a historical proposal; its claims are not new validation.