3 ms·
> In the future we might try doing some text classification to predict which applications will be good. In theory, that sounds a bit dangerous. If there was a
by mayank 16y ago
> In the future we might try doing some text classification to predict which applications will be good.
In theory, that sounds a bit dangerous. If there was a good model for predicting "success" from application text, then (a) the application can be gamed, and (b) you'd probably be picking up "hot" areas that reflect the business environment at the time more than the founders. YC seems to be more about humans, which makes a strong case for using only human judgement.
- ecuzzillo 16y agoIf they started using a text classifier instead of their brains to decide who to interview, they would be a) stupid and b) quickly out of money. a) and b) seem both overpoweringly unlikely, so I think more likely is that they'll just look at the results to see if there's anything interesting about them. (Wouldn't you like to see the results, just to see them? I would, and I have no intention of applying to YC anytime soon.) Initially I didn't understand why one would think they would be so stupid, and then I remembered what almost every other entity that receives applications is like.
- mayank 16y ago> Wouldn't you like to see the results, just to see them? Perhaps as indicators of the business environment. For example, perhaps your model would indicate that successful startups from the last few years would include the word "social" in them. I doubt the implication goes the other way.