Artificial Intelligence Concerns Tech Execs

Good Monday morning. It’s February 25th. Meteorological spring begins Friday.

Thank you, George Mason University’s Marketing 491 class and Professor Shaun Dakin, for allowing me to talk with them Thursday at the Fairfax campus. I always learn a lot from these sessions, and last week was no exception. Here’s an overview of what we discussed.

Today’s Spotlight takes about 5 minutes to read.

News to Know Now

The Open AI initiative announced it would not release its recent text generator project because “it looked pretty darn real.” We unpack the important story below.

Facebook is in new hot water over apps that report user health information to the company. It’s a day that’s name ends with the letter “y” – you expected good Facebook news?

Those Amazon Scout delivery robots being tested near Seattle have some cousins in college. George Mason University now has a fleet of 25 delivery robots bringing food including pizza, doughnuts, and coffee to people on campus. DoorDash has also tested the Starship Tech robots. The rollout video from the Mason launch is below.

Artificial Intelligence News

An algorithm’s writing reached a level that caused scientists to block its release. The text generator was reviewed by journalists including reporters from Axios and Wired. Fake political stories quoting real people and even fake Amazon reviews were well within the program’s grasp.

An excerpt from a story Axios had the AI write,

“China uses new and innovative methods to enable its advanced military technology to proliferate around the world, particularly to countries with which we have strategic partnerships,” the Pentagon said in its five-page strategy outline last week.The new U.S. strategy will be a major component of the White House’s first National Security Strategy, coming in two parts in September.” (Axios, Wired)

Visual disinformation, often called “deep fakes”, are also making news after software engineer Philip Wang created a website called This Person Does Not Exist. The site creates an entirely new image of a person from an algorithm every time a browser refreshes the screen. The very real image is of a person that never existed. Wang said that he created the site to show people how A.I. has advanced and its possibilities in today’s world. (Inverse)

CAPTCHAs are the small programs websites use to separate human visitors from automated traffic. You’ve seen them before–click this box to prove you’re a human or maybe select the parts of a picture showing a street sign. But machine learning and artificial intelligence is solving each new type of puzzle faster than before. “Machine learning is now about as good as humans at basic text, image, and voice recognition tasks,” said scientist Jason Polakis. (The Verge)

Tech Privacy News

“What’s the big deal if consumer tech has privacy problems?” a GMU student asked me. “They’ll just use the info to market more appropriate things to me.”

And he’s right.

But as Professor Dakin and I explained, there are many abuses from redlining based on race or gender to employment and other discrimination. NATO researchers have another take.

They used publicly available data against the participants in a military exercise in an Allied country. Their goals were to get service personnel to leave position, not fulfill their duty, or other actions. They took three to four weeks of prep time, used only information that the public could retrieve, and quickly identified the individual soldiers involved. Then they lured soldiers into fake social media groups and conversations, gained the exact locations of battalions as well as dates and troop movement information.

One conclusion: “The privacy features and settings…cannot be trusted not to leak information.” Our takeaway: people adept at pretexting and Internet research are not limited to NATO. Your organization can be targeted. Read the Stratcom report.

This wasn’t Facebook’s fault but the British government would like you to know that the company “intentionally and knowingly violated both data privacy and anti-competition laws” in their country. We’ve been telling you for months about the legal clashes between EU countries and Facebook, especially the United Kingdom and Germany. The report from Parliament is harsh and could have ramifications beyond that country’s nearly 40 million users. The U.S. government is also reportedly in talks with Facebook regarding a multi-billion dollar fine for the company’s data privacy practices.

Data privacy’s big news this week centered around an explosive WSJ expose that cites medical and other private data being sent to Facebook from mobile apps. The Journal has been following this story for months. This latest expose is going to spell trouble for the social media company for decades. WSJ and privacy researchers found:

Instant Heart Rate sends a user’s heart rate information to Facebook
Flo Period & Ovulation Tracker sends dates of a woman’s period or if she’s trying to get pregnant
Realtor.com sends locations and prices of homes users review
Better Me Weight Loss Workouts sends height and weight information.
Lose It also sent height and weight information along with exercise information, including sexual activity that the user entered.
Glucose Buddy sent blood pressure information and meal detail from food tracking.
Trulia sends zip code and some neighborhood information about homes users check.

Here is the most important part about these apps: they were downloaded from Google’s Play Store and Apple’s iOS App Store. They were not Facebook apps, but Facebook had reached seemingly legal data agreements with them to purchase information. And in some cases, the user didn’t have a Facebook account.When apps specify that they are going to share user information only with its trusted partners, those partners are often trusted as the result of writing a check.

The original WSJ article is here but requires a subscription. A free CNBC article is here.

Also in the Spotlight

Worth your time this week:

Advertisers are leaving YouTube again over “soft core” child exploitation and porn videos and comments left on them. The company has been deleting users (including commenters) and videos. AT&T and Walt Disney are among the companies that are boycotting the company.
Pinterest has filed for an IPO according to multiple reports, but as a direct listing. That means much less fanfare. Slow, steady Pinterest is winning praise for solid growth and banning anti-vaccination and other health-related disinformation.
40% of LinkedIn users check the site daily according to Search Engine Journal. The site now has 610 million members.