The Big Story

CEOs vs. Congress, round four

Emily Birnbaum writes: Mark Zuckerberg, Sundar Pichai and Jack Dorsey will appear before Congress today. If that sentence gives you deja vu, it's because this is the fourth time Zuckerberg has appeared before Congress in the last year, and the third for Dorsey and Pichai.

But this time, their appearance has new gravity. They're coming before the House Energy and Commerce Committee, which has jurisdiction over tech issues, and it's the first time they've appeared since Democrats swept the Senate and White House. It's also their first appearance since the Capitol riots, and the hearing is titled "Disinformation Nation: Social Media's Role in Promoting Extremism and Misinformation." So expect a lot of talk about Jan. 6. And a lot of fireworks.

Here are a few highlights from the CEOs' written testimony, to give you an idea of what's to come:

• Zuckerberg will get specific in his calls for Section 230 reform. He's basically calling for the most Facebook-friendly parts of the EARN IT and PACT Acts: He wants a law that requires social media companies to adhere to a set of "best practices" around removing unlawful content and forces them to publicize how they make content moderation decisions. Interesting that Facebook is one of the best-positioned companies to handle a law like that!
• Dorsey will hammer home Twitter's policy framework on content moderation: transparency, procedural fairness, algorithmic choice and privacy. Twitter's head of U.S. public policy Lauren Culbertson told Protocol during an event this week that Twitter's main focus is ensuring any Section 230 reform does not harm smaller companies while empowering the big guys.
• Pichai's testimony covers a lot of ground. He's going to talk about how YouTube handled the Jan. 6 riots, the role Google played during the 2020 election and even how Google promotes journalism. It's a reminder of just how many battles these companies are fighting simultaneously, and how interrelated each of them really is.

Lawmakers will have a lot of ideas for dealing with these issues, but not a lot of consensus. Aides have told me over the past week that the preparation for this hearing is almost nothing like last summer's hearing, which was meticulously organized. It's probably going to be a free-for-all. Buckle up!

Work

Another Slacklash

Slack had … a day yesterday. After launching its Connect DMs feature and being met with huge backlash from users, the company admitted it had made a mistake in the way it rolled out the product.

Harassment was the main concern about Connect DMs.Twitter's Menotti Minutillo pointed out how easy it was to harass people through Slack's DM invite system, and how hard it is to block those emails. That seemed to resonate with many users.

• "We are taking immediate steps to prevent this kind of abuse," Slack's Jonathan Prince said in a statement, "beginning with the removal today of the ability to customize a message when a user invites someone to Slack Connect DMs. Slack Connect's security features and robust administrative controls are a core part of its value both for individual users and their organizations."

The lesson here is simple: You should always investigate every pixel and action of a product by asking what a bad actor might do with it. ("They might say horrible things in their messages, and those messages would be basically unblockable" should have been an easy one to catch.)

• Better yet, do that investigation alongside the people who are most likely to be victimized, who have unfortunately seen it all already and can help you see it coming.

Too many tech companies have run into trouble over the years by believing only good people will use their products and only in the intended way, or by assuming that a curated group of beta testers will use the product exactly the same way as the entire teeming mass of the internet. "What's the worst that could happen" should be part of every company's product cycle. It certainly takes longer to backtrack after the fact.

AI

Translation for everyone!

Anna Kramer writes: Speech-to-text works well in English, but in other languages, it's hit or miss. For "minor" or lesser-used languages, it's often basically useless. This week, the community at Hugging Face is trying to fix that.

• They've organized a week-long sprint to take the open-source models for speech-to-text and massively improve them for these minor languages, using just volunteers.
• More than 200 machine-learning engineers and language experts all across the world this week have so far uploaded and begun revising more than 30 models for about 20 different languages, including Breton, Welsh and Turkish.

Don't underestimate what Hugging Face could do for speech-to-text. At its roots, the company is an open-source AI community, hosting and running hundreds of open-source ML and AI models — and it has more clout than you might imagine.

• The company started as an AI-powered chatbot, but quickly pivoted to become an open-source platform, Clément Delangue, the co-founder and CEO, told me. "It allows researchers to share models, version their models, run the inference of their models, both on Github and on their own product," he said.
• Hundreds of companies, including big-shots like Microsoft Bing and Grammarly, rely on Hugging Face ML tools to operate their software. If you're using BERT or GPT-3, you're using Hugging Face and you just probably don't know it.

And this is actually part of something much bigger. Large AI models can be wrecked by bad inputs, and in the best case are still hard to study because they take a ton of processing power and storage, and tech companies own many of the biggest and newest. Hugging Face is trying to change that by pushing to make these models more available, accessible and equitable, mostly through the work of community volunteers.

• Hugging Face has another highly ambitious project ramping up in the coming months as well: an academic partnership with over 150 institutions to build an extremely large open-source language model. "It's going to be a very long project," Delangue said, "because it's very much a science project. It might take six months to a year, and it might not even end up working at all. Maybe it's going to be a complete failure," he said. But he thinks it's worth trying, if it can help get the best stuff in everyone's hands.

A MESSAGE FROM AMAZON

A new study conducted by Amazon/Ipsos found that most Americans support raising the minimum wage. Results from the study, which involved interviews with more than 6,000 adults, revealed that two out of three support increasing the minimum wage to $15 an hour.

Read more

People Are Talking

On Protocol | Enterprise: AWS and Azure are great for huge companies but small businesses need something else, DigitalOcean CEO Yancey Spruill said:

• "Our founders ... nailed the fact that small businesses, startups [and] entrepreneurs have historically been underserved by big technology solutions. They need a simple, narrow and easy set of solutions."

Quilt's Ashley Sumner said she worries about phrases like "female founder" and "girl boss":

• "I worry that it allows investors to see founders who are women as a separate class from the rest of the founders. I worry it allows investors to write women founders smaller checks."

On Protocol | Enterprise: Improving the experience for neurodiverse employees starts in the interview, Microsoft's Serena Schaefer said:

• "With neurodiversity, some people think very fast. But some people may need a little more time to come up with the answer. Microsoft allowed everyone in the spectrum and accommodated everything they might need. It should be the standard for hiring for any candidate, not just those who happen to have a diagnosis."

• "I often say we are the Bernie Sanders of employers, but that's not quite right because we actually deliver a progressive workplace for our constituents: a $15 minimum wage, health care from day one, career progression, and a safe and inclusive work environment."
• Meanwhile, Amazon's communications team has adopted a similarly strange, similarly antagonisticTwitter strategy. It's a move, I guess.

Makers are the future of everything, Satya Nadella said:

• "Creation, creation, creation — the next 10 years is going to be as much about creation as it is about consumption and about the community around it, so it's not creating alone."

Making Moves

Cory Scott is the new CISO at Confluent, joining from Google. Melanie Vinson is also joining as chief legal officer, after a stint at Adaptive Insights.

Shou Zi Chew is ByteDance's new CFO, joining from Xiaomi.

Reddit fired an employee and promised to work on "evolving a number of relevant internal policies," after some users and moderators made their subreddits private in protest of her employment.

In Other News

• On Protocol | Policy: Facebook Messenger played a role in the Capitol riots. New evidence showed that in private Facebook messages, an Oath Keepers leader outlined explicit plans for an "insurrection."
• Trump's talking to FreeSpace about making a social network, Axios reported. The relatively unknown platform says it's "backed by science to positively reinforce good habits and make the world a better place."
• Arizona's Senate skipped a vote on the HB2005 bill, which would have allowed app developers to use third-party payment systems. Apple and Google had been lobbying hard against the bill.
• Google isn't running FLoC trials in Europe due to GDPR concerns, though it said it's "working to begin testing in Europe as soon as possible."
• The SEC is investigating SoftBank, seemingly over its suspected role as the "Nasdaq whale."
• The SEC said it would start enforcing a law that lets it delist Chinese companies from U.S. stock markets if they don't comply with U.S. auditing standards. Baidu, Alibaba and JD.com shares all fell on the news.
• On Protocol | China: China's women pro gamers are taking on the haters. But deep-seated misogyny, low pay and a belief women can't play are creating a vicious cycle that keeps talent away.
• India ordered an antitrust investigation into WhatsApp's privacy policy changes, raising concerns around the "unilaterally dictated 'take-it-or-leave-it' terms by a dominant messaging platform."

One More Thing

NFT of the day

It's digital art, created by a robot. It's "Sophia Instantiation," by Sophia the robot and the artist Andrea Bonaceto. It's the inevitable all-digital future of everything, and it sold yesterday for $688,888.00. Pretty soon they won't need us humans for anything ... except to buy stuff, I guess.

Update 3/24: This newsletter has been updated to properly spell Yancey Spruill's name and refer to DigitalOcean correctly.

A MESSAGE FROM AMAZON

A new study conducted by Amazon/Ipsos found that most Americans support raising the minimum wage. Results from the study, which involved interviews with more than 6,000 adults, revealed that two out of three support increasing the minimum wage to $15 an hour.

Read more