Copy
Want to be a campus rep? Reply to our email

July 24, 2020

 

 👋 <<First Name>>,

Biden is beating Trump in key battleground states (source: Fox News 😮) Yeah yeah, polls aren't always so good in predicting outcomes. But Trump may be a bit worried, given that he cancelled the Republican Convention in Florida.

THE STORIES 

💸 Less money, more problems

🤖 Humanoid intelligence

ECONOMICS

💸 Less money, more problems

 

WHAT 

  • The lockdowns and social distancing measures imposed across the world have changed consumption patterns considerably. For example, spending on restaurants, recreation, and transportation have decreased while spending on groceries and at-home entertainment have increased.

  • Although consumption patterns have changed, the basket of goods the US Bureau of Labor Statistics (BLS) uses to measure inflation has not been updated to reflected these changing consumption patterns.

  • Therefore, a new paper suggests that the official BLS inflation rate is a biased estimate of the economy’s true inflation. After creating a “Covid Basket”, the paper finds that both headline and core inflation are higher than official numbers indicate.

WHO

  • Alberto Cavallo is a professor at Harvard Business School, a Faculty Research Fellow at the National Bureau of Economic Research, and a member of the Technical Advisory Committee of the US Bureau of Labor Statistics. His work mostly focuses on inflation and its implications for economic policy.

WHY SHOULD I CARE 

  • Higher inflation means consumers are poorer because their same money can now buy less flamin’ hot cheetos. This effect is unfortunately amplified for poorer households because they have less $$$.

  • So what? 16 million Americans are still unemployed. Weekly unemployment claims shot up for the first time since March. The $600 unemployment boost and other economic relief programs are set to expire next week. Now, poor households are even poorer. And yet, the Nasdaq hit an all-time high on Monday. 

TL;DR 

  • Social distancing and lockdowns have changed consumption patterns, so the traditional basket of goods used to measure inflation gives a biased, lower estimate of the true inflation in the economy.

READ MORE HERE

Curious about how a low-income worker will be affected by the end of $600/week?
Look at this interactive chart.  


COMPUTER SCIENCE

🤖 Humanoid intelligence

 

WHAT 

  • The field of AI has a long history of trying to imitate human intelligence; neural networks are an attempt at mimicking billions of neurons in the brain, and reinforcement learning (RL) leverages how we learn proper actions from reward/punishment.

  • While there is obviously still a long way to go to simulating human intelligence, one gap is being tackled by some of the best RL experts in the field -- the same people who have built AIs to beat the world’s best chess and Go players.

  • That gap is a concept from psychology known as affordance theory, the notion that intelligent beings have an intuitive understanding of how their environment can be used. For example, we look at a tree stump and know that it can be used for sitting i.e. it “affords” that possibility.

  • AIs often don’t have that advantage. Instead, they have to try every configuration of an action on an object until they recognize which ones likely produce positive or negative outcomes. This process makes training these AIs time-consuming.

  • DeepMind has performed early experiments in which they granted an AI basic affordance intelligence in a very simple simulated environment. As hypothesized, training time to complete simple tasks in that environment sped up dramatically.

WHO

  • DeepMind, an AI company and research lab acquired by Google in 2014. They made headlines a few years ago after developing AlphaGo, a program which defeated the best Go player in the world and was the subject of an award-winning Netflix documentary.

WHY SHOULD I CARE 

  • We often like to think about a near future in which artificial intelligence is close enough to human intelligence to handle much of our mundane work. In reality, we are pretty far away from that, due in part to how long it would take to train different AIs for each specific task.

  • If an AI could understand how to interact with its environment from the outset, and could generalize that affordance intelligence to other environments, training could speed up dramatically.

TL;DR 

  • DeepMind from Google is performing the first experiments to show how the psychological concept of affordance can improve training time for their famous AIs

READ MORE HERE
 

PROFESSOR'S CORNER

Hope you guys liked our horse brainteaser on Wednesday. For today, we're taking another brainteaser inspired by the infamous STAT 110 course at Harvard. Joe Blitz is ridiculously smart (I mean, look at this guy)

  • A Russian gangster kidnaps you. He puts two bullets in consecutive order in an empty six-round revolver, spins it, points it at your head and shoots. *click* You’re still alive. He then asks you, “do you want me to spin it again and fire or pull the trigger again right away?”
What should you request? Should he spin it again for pull the trigger?

Recommended by Alison C., a former Stat-110er and one of our readers!
Check out the solution
Twitter
Facebook
Website
Copyright © 2020 The Office Hour, All rights reserved.


Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.