Aaron Schumacher retweeted
Turns out, you can learn a lot from a cheeseburger. 🍔📊
🎙️On @TheOfficialMRLA's "What’s with the Pineapple?" podcast, DoorDash Economist Erica Blom explains how the Cheeseburger Index works — and what it reveals about Detroit.
“The cheeseburger index looks at the average cost of a classic diner meal… The national average is $18.58; Detroit is $12.90.”
And that’s not all:
📈 94% restaurant resilience
📈 21% new restaurant growth
Detroit stands out in our State of Local Commerce report. ⤵️
🎧 Full conversation: bit.ly/4rGwFvL
Aaron Schumacher retweeted
🥚🥛Good morning, lower prices.
From DoorDash's first-ever State of Local Commerce Report, our Breakfast Basics Index — tracking the cost of breakfast grocery goods — is down 14% since March.
That’s one sign of strength in local economies: consumers are stretching their dollars further, and Main Street is proving resilient.
Dive into more insights: doordash-local-commerce.com/
Aaron Schumacher retweeted
deception rate being higher but shown lower is the funniest thing i have ever seen
Aaron Schumacher retweeted
YOU belong in tech! #HackDiversity = FREE 9-month career development program for early-career Black, Latinx, or underrepresented tech talent in #softwareengineering, #dataanalytics, and #MechE. 💻🚀
Applications are now LIVE. bit.ly/hack-apply2024
Aaron Schumacher retweeted
Neat visualization of Sweden life expectancy data by @planarrowspace revealing how there was no central tendency throughout history. Data in 1700 – 1900 is skewed by child mortality, where about half of kids died during their childhood, bringing the average to 35 years.
Aaron Schumacher retweeted
"Stay alert for investment scams involving cryptocurrency"
...
If it involves cryptocurrency, it's a scam.
Aaron Schumacher retweeted
Neural networks often pack many unrelated concepts into a single neuron – a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. In our latest work, we build toy models where the origins of polysemanticity can be fully understood.
Aaron Schumacher retweeted
Really excited about our new work:
an intuitive extension of decision trees to tiny tree sums 🌳 ➡️ 🌳+🌳 yields surprisingly large improvements in prediction performance & disentanglement, all without sacrificing interpretability arxiv.org/abs/2201.11931
Aaron Schumacher retweeted
Do we all agree that whiteboard coding interviews simply test for (high IQ + high motivation + ability to work evenings/weekends) and not actual coding ability?
If so, why pretend? Just tell candidates to improve their chess Elo rating by X% and hire them on the spot.
Aaron Schumacher retweeted
Excited to share my new blog post about why generalization in RL is hard, and why our standard RL algorithms are fundamentally incompatible with generalizing well!
bair.berkeley.edu/blog/2021/…
A more visual learner? Check out the talk instead: youtu.be/mOuf2fvNAhY
Why is generalization hard in RL? Can "just adding more data" fix it? Turns out that in general, the answer is no. In a new blog post, @its_dibya discusses this question: bair.berkeley.edu/blog/2021/…
Trying to generalize induces a POMDP, even if the problem is an MDP!
Aaron Schumacher retweeted
Great thread on the initial findings from Baby's First Years project. IRP director @profkmag, researcher @CostanzoMolly, and affiliates Greg Duncan @UCIrvine @KimberlyGNoble @TeachersCollege @Columbia Sarah Halpern-Meekin @UWSoHE & @Gen_Pov are among researchers on this study.
NEW FINDING: Cash support for low-income families changes their infants' brain activity. The first study from our #BabysFirstYears randomized control trial of poverty reduction is out, in @PNASnews. pnas.org/content/119/5/e2115…
( 🧵)
Aaron Schumacher retweeted
Looks like HKU students and faculty are set to het a little sitting area where the #PillarofShame used to stand. I overheard saws and drills here the other day, here’s what’s in the works.
Aaron Schumacher retweeted
Moxie’s analysis of web3 is really, really good. I had no idea about a lot of this stuff. moxie.org/2022/01/07/web3-fi…
Aaron Schumacher retweeted
Confounding Variables xkcd.com/2560/ m.xkcd.com/2560/
Aaron Schumacher retweeted
Replication crisis update
"none of the 193 experiments were described in sufficient detail"
"of the 50 experiments from 23 papers that were [repeated], effect sizes were, on average, 85 percent lower than those reported in the original experiments."
sciencenews.org/article/canc…
Aaron Schumacher retweeted
"You can't become the I HAVE NO IDEA WHAT I'M DOING dog as a professional identity. Don't embrace being a copy-pasta programmer whose chief skill is looking up shit on the internet." world.hey.com/dhh/programmer…
Aaron Schumacher retweeted
Facebook was one of the first companies to deploy face recognition at scale. Today @meta is announcing that it is turning it off. #ResponsibleAI is often deciding when to use, or in this case, not to use, AI. nytimes.com/2021/11/02/techn…
Aaron Schumacher retweeted
3. The literary characters Gelert, Argos, Gaspode, and Tock are all what, most distinctively?
CONCERNED ABOUT YOUR ATTITUDE LATELY
#LL89 #bestwronganswers
Aaron Schumacher retweeted
Many people have asked me how I make my interactive dashboards. I am surprised that many data scientists do not know HoloViz. So, I wrote an article about my favorite Python viz tools ❤️ HoloViz ❤️
towardsdatascience.com/visua…





























