Could AI model help decide which startups to fund?

An interesting thought in my head: could a sequence-to-sequence RNN model decide whether a startup is fundable or not (since there is data regarding which startups were funded and also future indicator of success, so the model would learn).

1. AI system makes determination whether to fund startup.

2. One or Three YC partners would also make an independent determination as to whether to fund the startup.

3. Then a weighted average rating would be computed to arrive at a final decision. Weights would have to be determined based on how reliable the model is.

Who would test this out at YC? How would you improve on this?