Week 04/15 - 04/21
This is the first post of my weekly blog series on interesting things I’ve read/listened. I also try to keep myself accountable in terms of self education by including the What I am working on section at the end.
News
-
- TLDR: found no conclusive evidence of Trump admin’s collusion with Russia; showed instances of Trump clearly frustrated about the investigation and tried to manipulate the justice system in many ways
Verge’s coverage of Andrew Yang
- AY background: running for 2020 presidential election as a Dem.; has startup experience (founded VFA - incentize young college graduates to move to emerging cities besides usual millenial hubs)
- Center to his platform: claims UBI ($1000 for every >18 US citizen) to be solution to automation
- Related: he’s appeared on Joe Rogan’s podcast (2 hrs) - link as well as other (center) right media outlets; also CNN’s coverage of him is atrocious - link; will add his book to reading list: War on normal people
- Thoughts:
- profile of a stereotypically successful Asian in America (Philip Exeter, Brown, Columbia Law) but pretty unusual for a presidential candidate - a huge plus for Asian representation
- curious to see what his voter demographic will mostly comprise of - my guess: technophiles; Asian Americans dying for polical representation; Reddit, 4chan trolls - the alt-right/white nationalist crowd (would be interesting to see how interactions between these groups play out)
‘Clean’ Chinese restaurant Lucky Lee (owned by white Jewish couple in NY) slammed for being culturally appropriative link
- Related: Asian eating house opened by Gordan Ramsay in London face similar backlash link
- Thoughts:
- In what way can a chef open a restaurant with a menu featuring cuisines from another ethnicity without risking cultural appropriation or insensitivity?
- Obviously these chefs must learn the history and other nuances around the culture they want to serve dish from but how is that consideration reflected either in their menu or the way their restaurant is advertised/run?
Big ideas
- AI’s white guy problem
- AI bias in hiring women, perpetuating discrimination in housing and criminal justice shows up because of lack of diversity in tech workforce
- tech diversity efforts have failed because companies wrongly focus on hiring “women in tech” (instead of more general diversity in race, gender, etc.) and “fixing pipeline” instead of focusing on retaining workers from underrep. groups by addressing systematic challenges such as compensation gap, harassment and power imbalance
- “Tech companies are built—and tech products are designed—with a “fantasy belief” that they exist independently of the sexism, racism, and societal context around them”
ML/Data Science
-
- What it is: a neural network model that guesses a hidden word based on context (words appearing before and after); takes as input large text corpus and produces vector space where each unique word in corpus is assigned a vector in space such that related words are closer in this space compared to non-related words (e.g. carrot and orange are close compared to carrot and George Washington)
- Foundational NLP technique, used by big tech companies for various NLP-related products
- What is it used for: as feature engineering step to preprocess input for other ML models; for improving query search (find results for similar phrases)
- Interesting notes:
- in word2vec’s representation, king - male + female = queen
- doctor - male + female = nurse (sexist bias in datasets that word2vec is trained on)
- Note-to-self: write an in-depth blog post
-
- Guest bio: data scientist at KPMG, a consulting firm
- Key takeaways:
- econometrics (def?) is useful for data science (more generally econ background is helpful for building relationships with other coworkers across different departments - if you can bring ppl on board with your ideas using language they understand you increase your value as a data scientist)
- econometrics (def?) is useful for data science (more generally econ background is helpful for building relationships with other coworkers across different departments - if you can bring ppl on board with your ideas using language they understand you increase your value as a data scientist)
What I am working on
- Coursera: deeplearning.ai courses
- Reviewed Week 1 and 2 in Course 1
- Recreational reading: Min Jin Lee’s Pachinko