4 ms·
I don't think it needs to be free for over-represented demographics, just open source so that it is clear that the algorithm is designed in the best interests o
by userabchn 4y ago
I don't think it needs to be free for over-represented demographics, just open source so that it is clear that the algorithm is designed in the best interests of users.
I like Andrej Karpathy's (I think) idea of training something like GPT-3 on each person and using the derived personality to identify those that you would get on well with (if the objective of the site is long term relationships). I suspect that a large leap forward is possible with something like this, although the technology is probably not ready yet.
- jnk345u8dfg9hjk 4y agothat's a pretty cool utopic idea, but also scary so we're just here on earth to be matched by an algorithm and live happily ever after? idk
- eklitzke 4y agoHow would you use the "derived personality" to identify those that you would get on well with? Even if you can come up with some kind of meaningful distance metric, it's definitely not clear that that's useful for matching individuals. For one thing it's completely ignoring the most important factor in dating, which is initial mutual physical attraction. For another, many (I would conjecture most) people are attracted to people who are quite different from themselves.
- runnerup 4y agoCould let users pick what type of relationship they're looking for: "forever-marriage", "summer fling", etc. For long-term marriage you could find couples to train a model for true positives. For the others you'd have to find couples who self-identify. Then determine not just "distance" but "vector"/"clustering" and identify which cluster combinations seem to be compatible for different types of relationships. This would also entrench present-day dating and marriage regimes. Then you could handle the physical attractiveness the same way Tinder does. Use the clustering model to pre-filter candidates, and present the user with photos for final thumbsup/down.