Eddie Lin

Great research on open-source by @Harvard: - $4.15B invested in open-source generates $8.8T of value for companies (aka $1 invested in open-source = $2,000 of value created) - Companies would need to spend 3.5 times more on software than they currently do if OSS did not exist I suspect that these numbers and impact are even greater for AI than for software (would be great to study!)

I added multi-modal (image, audio, video) support to my LLM command-line tool and Python library, so now you can use it to run all sorts of content through LLMs such as GPT-4o, Claude and Google Gemini simonwillison.net/2024/Oct/29/ll…

This year's conference in Chicago June 12-15 is one of the most exciting editions ever---with a huge array of different disciplines, real-world perspectives, and policy-oriented work that meets the stakes of this present moment. Register now: facctconference.org/2023/registrat…

@bernhardsson repetitive practices is probably right but would hardly agree when people say if you play music you are good at math. The real math only comes in when you're into music theory and composition which you can still play good instrument without. jazz cats are probably good at math.

There’s definitely a disproportionate amount of developers who are good at playing a musical instrument but I wonder what the causal graph looks like? I could see one thing driving the other but I could also see “likes to hyperfocus on repetitive things” as a huge confounder

The teams that build AI/ML systems for social problems should reflect the diversity & experiences of the communities the systems are designed to help. Help the Data Science for Social Good Summer Fellowship reach groups we need to train to achieve that dssgfellowship.org

@fulhack Luigi pipeline for large grid search!

What are the best startups or open source projects for hyperparameter optimization? (And question 2: what are people _actually_ using? sklearn.model_selection.GridSearchCV?)

There is not always a tradeoff between "accuracy" and "fairness" in ML/AI. Our recent empirical work investigating the accuracy cost of mitigating disparities across policy problems - we find that we can achieve fairness without sacrificing accuracy arxiv.org/abs/2012.02972

ML feature stores is another one of those systems that every other startup seems to reinvent in-house. Doordash seems to do have done a good job with it (1M qps!) on top of Redis with a bunch of tricks doordash.engineering/2020/11/19/bui…

github.com/puneetsl/lotion this saves my life

"If you are not paying the product, you are the product"

Thursday 11am EST: Solve Summer Academy on Data Science for Social Good Projects: Implementation & Project handoff, featuring friends from DSSGxUK @turinginst, @GlobalGiving, Medic Mobile and Exemption Project! Don't miss this finale! Register here: eventbrite.com/e/solve-summer…

Our DSSGxUK chapter run by @turinginst has the summer program closing ceremony tomorrow Sep 3 at 11am Eastern - learn about projects on improving the safety of childcare providers with @ofstednews and reducing corruption with @WorldBank Register at turing.ac.uk/events/dssgx-u…

@ecsalomon I like Clapton and Barry Manilow when I was 20 tho.

Talking to a friend this morning about getting older: "I was honestly hoping that it would be more like, 'liking Pearl Jam makes me old,' and less like, “now that I am old, I like 90s Clapton and Barry Manilow'”

Worth revisiting this @FiveThirtyEight post: On March 16, 18 experts estimated US COVID-19 cases on March 29 (today!). ALL 18 underestimated the cases with #s ranging from 10K to 75K. It's 137K right now. Only 3/18 had that number within their 80% CIs. 53eig.ht/3a9tULd

Wow, can’t even utter “Taiwan” in the WHO? You should set politics aside in dealing with a pandemic. FYI @WHO, 450+ news reports from 40+ countries so far positively covered #Taiwan’s handling of #COVID19. These reports do not mistake us as part of China & #TaiwanCanHelp. JW

New Data@Urban! How people can and are using our Ed Data Portal to analyze representation in Economics (and other fields). #DataScience @urban_institute/using-the-education-data-portal-to-tackle-representation-in-economics-3364bb210368" target="_blank" rel="nofollow noopener">medium.com/@urban_institu…

Why outbreaks like coronavirus spread exponentially, and how to "flatten the curve" wapo.st/2IL6UWR

pretty neat gis dashboard of coronavirus outbreak visualization. gisanddata.maps.arcgis.com/apps/opsdashbo…

NYC’s use of surveillance technology has raised serious privacy & civil liberty concerns. @genmaicha____ testified in front of the @NYCCouncil in favor of passing the POST Act, to require NYPD to disclose the surveillance technologies it’s using. ainowinstitute.org/ainow-geneviev…

What better way to spend winter break than applying for the 2020 Data Science for Social Good Fellowship at @CarnegieMellon in Pittsburgh. Application deadline is Jan 31 dssgfellowship.org