kevin

No complaints here, happy to work, happy to laugh, happy to teach!

[kevin @k2mcb]

@NoahLCowan @NateSilver538 yes indeed, thanks for doing the combinatorics! ChatGPT also thinks 8/10 rainy days are better than 2/10 non-rainy days =) . . .

[kevin @k2mcb]

@NateSilver538 which incredibly ironically is actually slightly more likely then it raining for 8 out of 10 days

[kevin @k2mcb]

@NateSilver538 Better is way too subjective. I understand the theory that you would want your models to be correct at the same rate of their confidence. However, if the true rate of rain prediction is 80%, then you have a 10% chance of it raining every single day.

[kevin @k2mcb]

@RivermanMedia @NateSilver538 i don't see why this is wrong. it's technically 100% accuracy because it's a classification problem. the loss function is going to be smaller if it rains every day then if it rains 8 / 10 days.

[kevin @k2mcb]

@realGeorgeHotz Totally agree, and me and my NLP friends truly hope you succeed!