Founder & CEO @hellomohana | previously: AI research @MILAMontreal, @rllabmcgill, recommender systems @shopify | S1'16 alum @recursecenter | runner and yogi.
@siobhanmitchell yum! I've been rotating between cinnamon, ginger, cardamom, and sometimes rose water. I find that with these spices and flavors, it tastes great even without sugar.
I am hosting an OB/GYN (MD) who is both a fertility and general female hormone health expert, on the Huberman Lab podcast today. What would you like me to ask her?
@itsmewaldi@paulg How do you measure 'qualified' and how do you remove human biases in that measurement? Not so much of common sense question anymore..
@alexiskold Explainability is about understanding what components of the input are responsible for triggering a certain output in a model. And understanding how sensitive is the model to small deviations in the input to flip the model's outputs.
@alexiskold FWIW my goal was not to (wo)mansplain here. Rather I figured that you may be interested in hearing a different POV to your statements from an AI scientist.
@alexiskold AI models being compared to random functions simultaneously as being compared to if/else functions, is the contradiction.
AI models are not random functions. They are all about pattern detection (!!), which is why saying that there is no pattern is incorrect.
@alexiskold Understanding if/else patterns in AI models is more difficult due to their complicated non-linear nature. That said, AI models are still all about pattern recognition.
@alexiskold non-linearity vs explainability are two separate questions. Demanding explanation from AI models is a very valid thing to do, and has been an active area of research in AI for a **very** long time.
@hubermanlab "Optimize" to me suggests that something is wrong with the individual right now. It's a word that's quite coupled with shame culture. Instead, why not use "grow", "work on your best version", or "improve"