Ian Vanagas

You should be able to instantly share a browsing session with somebody and invite them into the rabbit hole.

Gamifying of apps means companies can't make products look better, so they are tapping into your human biology and psychology.

3) What loss supervises the output predictions? What assumptions about the world does this particular objective make?

As a writer, looking for evidence of readership and engagement on Twitter makes you into the drunk looking for your lost keys under the street light.

"I watch the data all day long… I see every uptick or downtick. I put a video out, I watch how it trends. And then I'll make another video and see if that trends similarly." Then, she privates the videos that don't do well. She estimates she's made more than 20,000 TikToks, only a third of which are still public.

ML models are formed from combining biases and data. Sometimes the biases are strong, other times they are weak. To make a model generalize better, you need to add more biases or add more unbiased data. There is no free lunch.

By thinking about a ML problem first as a set of inputs and desired outputs, you can reason whether the input is even sufficient to predict the output.