meditation

Every ML Problem is a Recommendation Problem

I heard this line from someone who’s very dedicated at recommendation models, from paper reading group, an event I missed for couple of times.

• This mindset even lead to the fact that if we want to sell anything on website with LLM, it’s not LLM problem but recommender problem. There should be pair wise learning to help switching bettwen the products.
• Also law of showing up is awesome, you never know what you would learn, who you would know, what nuances you might exchange. TBH, in large chances the paper itself might suck, everything else and the cohort is awesome.

EE not drawing

EE not drawing again, chilling down the excitement hyped by my wife.

Keep high attention on a failing government and hope for luck is a very distracting thing. Join other people to watch the drawing is like having team mates buying lottery together. It won’t contribute to your winning, everyone just wish they can win for themselves, and the implication of winning is questionable by itself.

Played with Vectara

It’s far from an ideal vector database, I should have played with this when company has settled the choice on vector db.

I can see how it’s attractive to many of engineer doesn’t work with machine learning.

Again, if you don’t want to be the one who’s responsible for the choice, your voice won’t be or shouldn’t be as heard.