In September 2020, PhD student Colin Madland tweeted about how Twitter’s image-cropping algorithm favored his face—Madland is white—over that of his Black colleague.

Madland’s tweets racked up tens of thousands of likes and retweets and sparked a deluge of citizen science, as users tested how the algorithm cropped people of different races, genders, ages, and more.

Nine months later...Twitter announced the industry’s first-ever “algorithmic bias bounty challenge,” to encourage—and reward—more formal investigation of the very same image-cropping algorithm. Though not a silver-bullet solution, it’s rare for major tech companies to allow such direct, public engagement with the algorithms that govern their platforms.

• The image-cropping algorithm Twitter put on display was partly decommissioned three months before the competition, but it had been deployed since 2018 to all ~200 million of Twitter’s daily users.

So, how’d it all come together?

To pull off this competition, the META team had to 1) get the sign-off from Twitter CTO Parag Agrawal 2) move fast—they wanted it to be part of DEF CON 2021’s AI Village, and the megaconference was rapidly approaching—and 3) figure out how to create a rubric to grade AI harms. They also had to run it all by CEO Jack Dorsey.

“There’s a moment, and then an idea happens, and you have to actually be ready—like the world has to be ready, in a sense,” Rumman Chowdhury, director of the META team and a driving force behind the bias bounty, told us. “In our cases, the field had to be ready. I think three years ago...if someone launched a bias bounty, no one would’ve understood what to do with it. We needed to be at a place where it needed to happen.”

First prize ultimately went to Bogdan Kulynych, a researcher at Switzerland’s EPFL technical university, who found that the algorithm prioritized young, slim, and lighter-skinned faces.

• Other winners found that the algorithm rarely chose to feature people pictured in wheelchairs, preferred emojis with lighter skin tones, and was more likely to showcase memes with English text than Arabic script.

Looking ahead...Chowdhury told us she sees this transforming from a competition into an ongoing vulnerability rewards program at Twitter. But, she acknowledged, that would require a lot of investment.

Big picture: Sasha Costanza-Chock, director of research and design at the Algorithmic Justice League (AJL), told us that while they’re enthusiastic about any sort of initiative to shine a light on bias and harms in AI systems, bounty programs are just a small step forward.

Click here to read the full piece on how the bounty program came together.—HF

Soon, the automaking world may care as much about recycling as your new agey family member. Specifically, battery recycling—with EV adoption set to surge in the coming years, crucial metals like lithium, cobalt, copper, and nickel could be in short supply if companies rely on new mining alone.

Case in point: Last week, Ford invested $50 million in Redwood Materials, an EV battery recycling company led by a cofounder, and former CTO, of Tesla. As part of the investment, Redwood will help Ford recycle its own EV battery materials and create a “closed loop” supply chain.

• The battery recycling company claims it can recover 95% of the elements used in batteries, and it also closed a $700 million funding round in July.
• It’s partnered with Panasonic and Amazon to recycle lithium-ion batteries and other e-waste in the past.

Battery recycling is key to building a sustainable all-electric future, experts say, both because the supply of necessary metals is finite and because of the inhumane conditions under which certain minerals, like cobalt, are often mined.

• Cobalt and nickel are more at risk of shortages than lithium, but even that relatively abundant element will face supply issues if EV adoption scales as rapidly as is expected.
• Companies are also attacking the opposite end of these issues, attempting to create novel (e.g., seafloor mining), more efficient, and/or safer forms of mining.

But, but, but...For now, it’s still cheaper to make batteries from newly mined materials than recycled ones. Experts expect the economics to change as the Fords of the world begin pumping out EVs, as has happened with the much less valuable lead-acid batteries used in gas-powered cars: 98% of lead-acid batteries are currently recovered or recycled.—DM

Folks, we’ve said it many times, and now we’re saying it again, for the last time: We published a ton of crypto content across Morning Brew in August, and we really think you should check it out.

We’ve already highlighted Emerging Tech Brew’s stories, so this time we’ll spotlight a few of the pieces published by our friends over at the flagship newsletter:

• 7 charts that explain the current state of crypto
• Should you invest in crypto?
• An interview with a crypto crime fighter

For more...You can read the rest of our crypto stories at this hub.—DM

Stat: Legal tech companies have already attracted more than $1 billion in funding this year, breaking the previous full-year record of $989 million as the slow-moving industry digitizes.

Quote: “The more wireless technology is embedded in our lives, the more complicated these transitions become.”—Harold Feld, a senior vice president of Public Knowledge, re: phasing out 3G coverage

Read: Deep learning’s diminishing returns.

It adds up: Multiple deadlines + pressure for accuracy = stressed accountants. That’s why FloQast’s software was built to help accounting teams operate more efficiently. As former accountants themselves, they’ve developed a few interesting resources to help you decompress after closing the books. Learn more here.*

^{*This is sponsored advertising content}

Tuesday: Amazon is hosting its fall hardware event. Details are scant right now, but the company has previously used the event to both release new versions of old standards (e.g., updated Echo speakers) and to unveil hardware projects, like its Ring–branded in-home drone.

Wednesday: The US and EU will kick off a brand-new trade and tech council in Pittsburgh. Working groups include: tech standards, climate and green tech, and data governance. Let’s hope they avoid the topic of French submarines.

Thursday: The House of Representatives will vote on the $1 trillion bipartisan infrastructure bill this Thursday, which includes tens of billions in funding for EV charging stations, electric buses, broadband, and smart-grid projects, among other critical infrastructure initiatives. Like functioning bridges.

Researchers from the Lincoln Agri-Robotics center have created a data set of ~3.5K images of strawberries.

...Why? The aim is to train deep learning models to detect, from visual data alone, whether a piece of fruit is ripe or unripe. The data set was created via what the researchers say is a “novel data augmentation technique,” i.e., a new, color-based way to systematically create slightly altered copies of real images in order to build a larger data set.

The researchers say their experiments suggest color-based data augmentation is more useful for detecting the ripeness of fruit than structure-based augmentation, a common and preexisting technique.

Catch up on the top Emerging Tech Brew stories from the past few editions:

• Growing medical cannabis? You’ll need some tech essentials first
• Spaced Ventures aims to be the Kickstarter of early-stage space startups
• Smart grids could soften the blow of cyberattacks, but make them more common