A Blog by Jonathan Low


Oct 25, 2020

US FDA May Not Issue Emergency Covid Vaccine Authorizations

There is growing concern that the emergency use authorizations could be abused for political reasons, potentially ruining the opportunity to see which vaccines are most effective. 

Granting expanded access as an alternative would limit both uses and claims of success as the drugs' real world efficacy  are investigated. JL

Helen Branswell reports in Stat:

The Food and Drug Administration is getting cold feet over issuing emergency use authorizations to allow for the widespread early deployment of Covid-19 vaccines. Instead, the agency may be explore using expanded access, a more limited program used for investigational drugs. The fear is that early authorization of vaccines could squander a one-time chance to determine how well the various vaccines work and which work best in whom.

How Ethical Hackers Are Trying To Protect the US Election

Electronic voting machines are just the tip of the iceberg. The entire system of tabulating, storing and reporting information is at risk. 

A group of self-organized cybersecurity experts are quietly trying to help secure the system behind the scenes. JL

Chris O'Brien reports in Venture Beat:

While public fears focused on electronic voting machines, the Department of Homeland Security defined the voting system as: “storage facilities, polling places, and centralized vote tabulation locations used to support the election process, information and communications technology to include voter registration databases, voting machines, and other systems to manage the election process and report results on behalf of state and local governments. “All software is vulnerable. It just depends on how long you’re taking to look to find those vulnerabilities. Humans write code, and humans make mistakes.”

AI May Be Able To Spot Incipient Alzheimers From Written Word Patterns

AI identified misspellings, repetition and grammatically simple phrasing reveal risk of Alzheimers with 70% accuracy. JL

Jeremy Hsu reports in Scientific American:

AI could screen for Alzheimer’s  by analyzing writing. A team from IBM and Pfizer says it has trained AI models to spot early signs of the illness by looking at linguistic patterns in word usage. The AI model was able to detect linguistic features that are related to early signs of cognitive impairment. They include misspellings, repeated words and the use of simplified phrases rather than grammatically complex sentences. The model achieved 70% accuracy in predicting which participants developed dementia associated with Alzheimer’s before age 85.

If Everyone Is Home Due To Covid, Why Are TV Sports Ratings Down?

After cancelling virtually every sport in the spring, all of them are now competing with each other in the fall. 

Oh, and then there's a presidential election as well as a pandemic. Plus, after being home so long, people are just burned out on screen watching. JL

Kevin Draper reports in the New York Times:

Ratings for the N.B.A. finals were down 49%, and the N.H.L.’s Stanley Cup finals were down a whopping 61%. Major League Baseball regular season and playoffs, US Open tennis, US Open golf, Kentucky Derby, Preakness and college football have all had ratings declines of at least 25% compared with 2019. Even the usually untouchable N.F.L. was down 13% through Week 5. Compared with 2019, total viewership across all television was down 9% in 2020, and 10% during prime time. (And) cable news in was up 79% compared with last year because the presidential election and pandemic.

Why GDP Growth May Have Less Political Impact This Pandemic Election Year

The drop in March and April was so catastrophic that even extraordinary growth cannot overcome its effects. And the fact that GDP growth has slowed in the past few months suggests the recovery is going to take longer. JL

David Wilcox reports in Vox:

Most economic forecasters expect the GDP growth figure announced on October 29 to be somewhere in the neighborhood of 25 to 30% at an annual rate. If the number is in that range, it will (be) the fastest GDP growth ever posted. But it will not mean the economy is booming, or that it has recovered (because of)  the sharpest collapse ever recorded during the second quarter, 31% at an annual rate. Indicators suggest momentum has slowed. In May and June, the pace of household spending rebounded quickly following collapse in the preceding two months. But in July and August, improvement slowed

How Covid Is Boosting Chinese Telemedicine, Which Could Become Global Model

Convenience, speed and cost are driving patients online just as they have with other consumer services. JL

Martin Mou reports in the Wall Street Journal:

For Chinese consumers, one of the biggest draws has been the use of messaging apps to communicate with doctors. Doctors can review previous medical tests, inspect photos of conditions and prescribe medicines for patients of chronic diseases, then communicate with patients via messaging apps, phone calls or video chat. Consultation fees  range from the yuan equivalent of a few U.S. dollars to several dozen depending on doctors’ credentials. WeDoctor has connected 7,200 hospitals across China to its online platform, which also allows users to book in-person appointments.

Oct 24, 2020

The Reason AI With Human Features May Be Less Trusted

Would you buy a used car from this robot? Research shows that people are prone to take simple recommendations from AI because they believe it is 'just an algorithm' and not designed to exploit them. 

But AI with human features or names may be considered less trustworthy because they are perceived to have intentions like humans. JL

TaeWoo Kim reports in Singularity Hub:

Research suggests people find AI’s recommendations more persuasive in situations where AI shows easy steps on how to build personalized health insurance or how to choose the right tennis racket, rather than why these are important to do in a human sense. People are much more likely to accept “unfair” offers if proposed by an AI because we don’t think an AI has a malicious intent to exploit, it’s just an algorithm. Giving AI human features or a name could mean people are more likely to believe an AI has free will (but) AI may be seen as having its own intentions, which could be exploitative.