Skip to main content

The ex-Google researcher staring down Big Tech

How race and identity are shaping politics, policy and power.
Mar 18, 2022 View in browser
 
The Recast header image

By Brakkton Booker

With help from Rishika Dugyala and Teresa Wiltz

Timnit Gebru

POLITICO illustration/Photo by Getty Images for TechCrunch

What up Recast family! Moderna seeks FDA authorization for a second booster for adults, an American civilian is killed in a Russian assault in Ukraine, and a Russian court extends WNBA star Brittney Griner's detention until at least May. Today, we're kicking things off with a thought leader in ethical issues in artificial intelligence. 

Timnit Gebru is a renowned computer scientist and leading expert on the racial bias of facial recognition tools.

She is also among the 40 people and organizations named to The Recast Power List we debuted last week. She made the list for her whistleblowing efforts to root out the harm technology can have on marginalized groups — not to mention her advocacy of greater diversity and inclusion within Silicon Valley.

Her controversial firing from Google in late 2020 jolted the tech industry. She says her termination was over academic research she co-wrote spelling out the dangers of some of the AI technology that is used to power Google's search engines.

"Of course, all of it — in my view — was related to the fact that I am a Black woman," she said when asked if race played a role in her firing during my POLITICO Live interview with her this week.

(In an email to Google employees, Jeff Dean, the head of Google AI, said Gebru's paper didn't "meet our bar for publication" and that Gebru required "that a number of conditions be met in order for her to continue working at Google. … We accept and respect her decision to resign from Google.")

Gebru is now charting a path of her own through her own organization, Distributed Artificial Intelligence Research Institute (DAIR). It aims to raise awareness about the harm AI technologies can cause while also pushing tech giants and lawmakers for greater oversight and accountability.

Gebru, who was born in Ethiopia to Eritrean parents, is committed to disseminating DAIR's research among communities impacted by harmful AI, particularly those in Africa.

We chop it up about her time at Google, what more lawmakers could be doing in their oversight of big tech firms and if she sees a difference in how the media covers women of color whistleblowers in tech versus their white counterparts.

◆◆◆

This interview has been adapted from a Politico Live event and edited for length and clarity.

THE RECAST: How much do you think the scrutiny you received at Google was tied to your identity as a woman of color challenging the status quo and challenging authority?

GEBRU: All of it. I mean, where do I even start?

So, the paper that we wrote, first of all, was from the perspective of people in marginalized groups. There were about seven authors … from the LGBTQ+ community, Latinos, Black people, women, members of the indigenous communities. We also had about four or five continents represented. So in this case, the conversation about large language models that we had was from that perspective.

And so [Google managers aren't ] used to hearing people express our opinions about potential harmful technology from our perspective.

I had been having so many issues at Google during my two years there. I started talking about the issues that I saw, especially with respect to women and Black people and how they were being treated. I started facing resistance right away. The latest episode that got me fired was just one of a string of issues that I was facing there.


 

logo test

Every week, we sit down with diverse and influential characters who are shaking up politics.

Who should we profile next? Let us know. Email us at therecast@politico.com.

 


THE RECAST: We didn't see the same kind of groundswell of support that we saw for, say, Frances Haugen, the Facebook whistleblower, who is a white woman. Haugen testified before Congress alleging Facebook magnifies hate and misinformation on its platform and had this trove of documents that she said backed up her claim.

What do you make of the difference between the support that Haugen received and what you received from mainstream media?

GEBRU: First of all, I want to say I am happy that she got that support. I am happy that she was able to have her voice heard in [mainstream outlets] like "60 Minutes." I hope that it results in tangible change.

That being said, I mean, we do know that Black women are not heard, right? When we speak up, we're "angry" — that's how we are characterized. And I'm not quite sure what kind of backlash she faced, but the backlash I faced was huge.

I still have people talking about how my work was "subpar."

Frances Haugen

Former Facebook employee Frances Haugen testifies before the Communications and Technology Subcommittee of House Energy and Commerce Committee in December. | Alex Wong/Getty Images

And in terms of "60 Minutes," there was even an episode on face recognition and wrongful arrests of people based on flawed technology. If I had to name one person who is a pioneer in this space, it's Joy Buolamwini. And they were talking about her work, basically, and they had a white man discuss her work.

So in short, it's no secret how we are treated when we speak up as Black women.

I'm grateful that Francis Haugen did speak up. Hopefully, the public is getting some of the message that we have to rein in these large tech companies. They have too much power, and they're doing a lot of harmful things without any sort of checks and balances or oversight on them.

[Editor's Note: Buolamwini says she spoke with "60 Minutes" producers for hours for their May 2021 story on facial recognition, including providing research recommendations and people to interview. In a statement, CBS said while they didn't interview her on camera "she was an important part of the research we did due to her work in this field."]

THE RECAST:  President Biden ran on a platform promising to root out inequities in federal agencies and programs. Has his administration done enough to tackle the issue of discriminatory AI?

GEBRU: I'm glad to see that some initiatives have started. I like that the Office Of Science And Technology Policy (OSTP), for instance, is filled with people I respect, like Alondra Nelson, who is now its head.

My biggest comment on this is that a lot of tech companies — all tech companies — actually, don't have to do any sort of test to prove that they're not putting out harmful products.

So, we were talking about face recognition a long time ago and … we don't have a federal bill. We even have all of these segments of the government using it: law enforcement and you saw ID.me with the IRS.

So the issue is we are looking to the government to regulate this technology and have some sort of oversight, but then we're also seeing the government using this unregulated harmful technology.

This is an issue for me.

I'd like to see the federal government come up with regulation, but also look at itself and what kinds of technology that it is using as well.

THE RECAST:   Where do you anticipate this action coming from? I see lawmakers relish the fact that they can bring big tech heads in and grill them on their technologies. But when it comes to passing oversight laws, there's little appetite for getting that legislation across the finish line.

GEBRU: There absolutely needs to be grassroots pressure.

But what I'm trying to say is that the grassroots people are very under-resourced and overburdened. They're constantly trying to prove harm.

The problem is that these tech companies have so many resources. They have people all day whose jobs are to lobby members of Congress and senators and other legislators, even internationally in the EU, etc.

People like me, I try to talk to as many people as I can in the government. But I don't have the resources or time to do this all day. So, that action has to come from everybody. I think everybody has a role to play.

THE RECAST: Well, let's talk DAIR. A lot of organizations say they want to kind of influence tech practices from the outside. But, as you mentioned, it's a resource issue. What makes DAIR distinct?

GEBRU: I really do believe in the phrase: power to the people. I think that people have power and we can exercise that power.

So, with DAIR, that is kind of the underlying philosophy that we have. We know we're not going to have the unlimited money that Google has, or even the military has.

Google San Francisco

Google San Francisco in January 2018. | Santiago Mejia/San Francisco Chronicle via AP

Historically, how AI research has been developed, there's been two major sources of funding: One is something to do with warfare, like DARPA (Defense Advanced Research Projects Agency). If you look at even self-driving cars, the Robo cops, etc., machine translation has roots in the Cold War. So it's either that, or these large corporations trying to maximize profit.

At DAIR, … we make sure that the money that we're getting, at least, is not motivated by either of these two goals. And, we want to be able to stop the harms of AI without being fired from places when we try to educate our community about the potential harms of AI.

We want to do the development and the research in a way that will benefit our communities. And, when we see existing technology that's harming our communities, we want to be able to work on helping stop that technology without being persecuted.

THE RECAST: Some of the push you're trying to do with DAIR is work with countries in Africa that can help kind of illuminate these problems with AI or find some potential solutions that could be applied globally?

GEBRU: Absolutely, we have to have an international view.

I am of Eritrean descent. I was born and raised in Ethiopia. What has been happening in Ethiopia?

There has been a war with rape used as a weapon of war, starvation used as a weapon of war, unspeakable horrors that you can imagine. An estimated 500,000 people dead in the last 16 months in a region of maybe 6 million people.

I have not heard a single news story about it. Eritrea has one of the highest refugees per capita and what happens to them? They get drowned in the Mediterranean. And so, there's all sorts of tech involved in this kind of stuff, too, that we don't have time to talk about right now.

These tech companies are concentrated in the U.S., that are impacting the entire world, but the entire world doesn't have the ability to impact them. So with DAIR, that's what we're hoping to do. I also don't want to contribute to the brain drain. If someone in an African country wants to work at DAIR, I don't want them to have to leave that community and then have to move to Silicon Valley or something like that. I want them to enrich their own communities. So, that is a part of the philosophy of DAIR as well.

Watch our live conversation with Timnit Gebru and Sen. Alex Padilla on changing power.

Recast live event video

THE RECAST: My final question here, what advice do you have for those who aspire to do what you are doing in shaking up these power dynamics in the tech industry?

GEBRU: Just remember: power to the people. People have power.

I think we forget how much power people have when they work collectively and when they have collective advocacy, when they do collective organizing. So, when you're thinking about what to do and what to change, don't think about it just as yourself, you as an individual.

I always think about myself in the context of the tech organizing movement. There are a lot of people who play a lot of roles. There are a lot of people behind the scenes, whose names you don't know about, that I talk to all the time. They're part of this movement [too].

So I always ask people to think about where you think you might fit in this movement. How you can build coalitions and advocacy, and solidarity. You advocate on someone else's behalf. Someone else advocates on your behalf. And that's kind of what creates change.

◆◆◆

OK fam, please hit us up at TheRecast@politico.com for tips, story ideas or folks you think would be intriguing Friday Q&A guests. Now to our weekend to-dos!

Author Cristina Rivera Garza lost her sister 30 years ago — a murder that remained unsolved. Last year, Rivera Garza's memoir, "El Invencible Verano de Liliana," published in Mexico, helped identify the suspect. Look for the English language edition, "Liliana's Invincible Summer: A Sister's Memoir," next year.

Also on our reading list: Eloghosa Osunde's new novel, "Vagabonds," set in the megalopolis of Lagos, where supernatural beings roam the streets, wreaking all kinds of havoc.

We're so pumped for next week's return of "Atlanta," Donald Glover's most excellent FX series. This time, he and his crew venture out of the ATL for Paper Boi's European tour. Expect all kinds of surreal craziness to ensue.

Also back next week: "Bridgerton," regrettably sans Regé-Jean Page.

And we're so ready to dive into "Pachinko," a juicy Korean epic spanning continents and generations.

Pachinko Trailer

Heading into the weekend with the dreamy vibes of Colombian singer/songwriter Camilo.

 Colombian singer/songwriter Camilo NPR

TikTok of the Day: The feeling after filling your tank.

TikTok day after filling your tank

 

Follow us on Twitter

Brakkton Booker @brakktonbooker

Rishika Dugyala @rishikadugyala

Teresa Wiltz @teresawiltz

 

Follow us

Follow us on Facebook Follow us on Twitter Follow us on Instagram Listen on Apple Podcast
 

To change your alert settings, please log in at https://www.politico.com/_login?base=https%3A%2F%2Fwww.politico.com/settings

This email was sent to rouf@idiot.cloudns.cc by: POLITICO, LLC 1000 Wilson Blvd. Arlington, VA, 22209, USA

Please click here and follow the steps to unsubscribe.

Comments

Popular Posts

📄 Liaqat Ali Ali shared Waheed Zaman's post

  See the post that he shared.           Facebook                 📄 Liaqat Ali Ali shared Waheed Zaman 's post. 26 April at 12:09   View               This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more      

📄 Sazzad Khan shared Shakil Khan's post

  Ludo, see the post that he shared.           Facebook                 📄 Sazzad Khan shared Shakil Khan 's post. 27 April at 21:42   View               This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more      

Afjal Hossain wants to be friends on Facebook

  1 mutual friend - Works at won businass - Atrai, Rājshāhi, Bangladesh - 3,883 friends - 6 photos - 3 groups           Facebook             Afjal Hossain wants to be friends with you on Facebook.   Afjal Hossain Works at won businass · Atrai, Rājshāhi, Bangladesh 1 mutual friend · 3,883 friends · 6 photos · 3 groups               Confirm request     See all requests             This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more.      

DeepMind Gemini AI could rival ChatGPT / Congress restricts ChatGPT use / Unity launches AI marketplace

Plus: C-suite leaders plan to hire due to generative AI Inside AI For June 27, 2023 Thank you to our sponsor Hey Insiders,  Welcome to Inside AI! Here are today's top stories: Google DeepMind is developing a Gemini chatbot to rival ChatGPT. 64% of C-suite leaders say they plan to increase hiring due to generative AI. U.S. House members were instructed to limit ChatGPT usage to the paid version. Scroll down for some quick hits, and then let us know what other content you'd like to see in future issues!  Beth p/beth-duckett 1 Google DeepMind, its AI research lab, is developing a new chatbot called Gemini that could rival or surpass OpenAI's ChatGPT, according to CEO Demis Hassabis. Gemini, built on the capabilities of DeepMind's AlphaGo AI system, could analyze text, plan, and solve problems, Hassabis told Wired. More: AlphaGo is DeepMind's AI system that defeated a professional human player i...

The $975 million TikTok mystery

Red flags pile up in Khaby Lame’s mega deal. View in browser | Subscribe ...

Breaking News: Dems fall short of flipping deep-red House seat in surprisingly competitive Tennessee special election

Dems fall short of flipping deep-red House seat in surprisingly competitive Tennessee special election Breaking News Alert ...

Start separating your leaves from your trash

Presented by Uber: POLITICO's must-read briefing informing the daily conversation among knowledgeable New Yorkers May 31, 2023 View in browser   By Danielle Muoio Dunn and Zachary Schermele Presented by ...

75% of software engineers faced retaliation after reporting wrongdoing / AWS announces new updates to CodeWhisperer at re:Invent event

Plus, Plane raises a $4M seed round Inside Dev For November 29, 2023 Thank you to our sponsor Here's a look at today's Dev briefing. Engprax's research discovered that 75% of developers who reported wrongdoings faced retaliation.  AWS unveiled new updates to CodeWhisperer at re:Invent.  Microsoft open-sourced Azure RTOS.  Thank you. Sudarshan p/sudarshan 1 A recent Engprax/Survation research revealed that 75% of software engineers in the U.K. faced retaliation after reporting wrongdoing to their employers. More than half (53%) of respondents said they suspected wrongdoing at work. More: Software engineers who did not report wrongdoing pointed to fear or retaliation from management (59%) and colleagues (44%) as the main reasons. The investigation also indicated that engineers feel unsafe to voice concerns because of flaws in the "industry standard" metrics frameworks used to asses softwar...

Trump verdict puts Illinois GOP on the attack

Shia Kapos' must-read rundown of political news in the Land of Lincoln May 31, 2024 View in browser   By Shia Kapos TGIF, Illinois . The 2024 presidential campaign just got a new talking point. TOP TALKER PLEDGING ALLEGIANCE: Illinois Republicans and a certain former gove...

Meta removes disinformation campaigns / Google Maps disables features in Ukraine / NIO seeks HK listing

Plus: OnePlus's upcoming phone can fully charge in 15 minutes. Inside.com Part of   Network February 28, 2022 Presented by Facebook parent company Meta said it took down two disinformation campaigns that targeted Ukraine and were spreading anti-Ukrainian propaganda. Russia-linked hacking group Ghostwriter, which perpetrated one of the campaigns, gained access to real Facebook accounts and posted YouTube videos "portraying Ukrainian troops as weak and surrendering to Russia," according to Meta. The company said it blocked phishing domains that Ghostwriter used to gain access to the accounts and beefed up security. More: Meta described the second disinformation campaign as "coordinated inauthentic behavior" made up of about 40 accounts, pages, and groups on Facebook and Instagram.  The campaign ran several websites disguised as independent...

Popular Posts

📄 Liaqat Ali Ali shared Waheed Zaman's post

  See the post that he shared.           Facebook                 📄 Liaqat Ali Ali shared Waheed Zaman 's post. 26 April at 12:09   View               This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more      

📄 Sazzad Khan shared Shakil Khan's post

  Ludo, see the post that he shared.           Facebook                 📄 Sazzad Khan shared Shakil Khan 's post. 27 April at 21:42   View               This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more      

Afjal Hossain wants to be friends on Facebook

  1 mutual friend - Works at won businass - Atrai, Rājshāhi, Bangladesh - 3,883 friends - 6 photos - 3 groups           Facebook             Afjal Hossain wants to be friends with you on Facebook.   Afjal Hossain Works at won businass · Atrai, Rājshāhi, Bangladesh 1 mutual friend · 3,883 friends · 6 photos · 3 groups               Confirm request     See all requests             This message was sent to ludomallam@idiot.cloudns.cc . If you don't want to receive these emails from Facebook in the future, please unsubscribe . Facebook, Inc., Attention: Community Support, 1 Facebook Way, Menlo Park, CA 94025         To help keep your account secure, please don't forward this email. Learn more.      

DeepMind Gemini AI could rival ChatGPT / Congress restricts ChatGPT use / Unity launches AI marketplace

Plus: C-suite leaders plan to hire due to generative AI Inside AI For June 27, 2023 Thank you to our sponsor Hey Insiders,  Welcome to Inside AI! Here are today's top stories: Google DeepMind is developing a Gemini chatbot to rival ChatGPT. 64% of C-suite leaders say they plan to increase hiring due to generative AI. U.S. House members were instructed to limit ChatGPT usage to the paid version. Scroll down for some quick hits, and then let us know what other content you'd like to see in future issues!  Beth p/beth-duckett 1 Google DeepMind, its AI research lab, is developing a new chatbot called Gemini that could rival or surpass OpenAI's ChatGPT, according to CEO Demis Hassabis. Gemini, built on the capabilities of DeepMind's AlphaGo AI system, could analyze text, plan, and solve problems, Hassabis told Wired. More: AlphaGo is DeepMind's AI system that defeated a professional human player i...

The $975 million TikTok mystery

Red flags pile up in Khaby Lame’s mega deal. View in browser | Subscribe ...

Breaking News: Dems fall short of flipping deep-red House seat in surprisingly competitive Tennessee special election

Dems fall short of flipping deep-red House seat in surprisingly competitive Tennessee special election Breaking News Alert ...

Start separating your leaves from your trash

Presented by Uber: POLITICO's must-read briefing informing the daily conversation among knowledgeable New Yorkers May 31, 2023 View in browser   By Danielle Muoio Dunn and Zachary Schermele Presented by ...

75% of software engineers faced retaliation after reporting wrongdoing / AWS announces new updates to CodeWhisperer at re:Invent event

Plus, Plane raises a $4M seed round Inside Dev For November 29, 2023 Thank you to our sponsor Here's a look at today's Dev briefing. Engprax's research discovered that 75% of developers who reported wrongdoings faced retaliation.  AWS unveiled new updates to CodeWhisperer at re:Invent.  Microsoft open-sourced Azure RTOS.  Thank you. Sudarshan p/sudarshan 1 A recent Engprax/Survation research revealed that 75% of software engineers in the U.K. faced retaliation after reporting wrongdoing to their employers. More than half (53%) of respondents said they suspected wrongdoing at work. More: Software engineers who did not report wrongdoing pointed to fear or retaliation from management (59%) and colleagues (44%) as the main reasons. The investigation also indicated that engineers feel unsafe to voice concerns because of flaws in the "industry standard" metrics frameworks used to asses softwar...

Trump verdict puts Illinois GOP on the attack

Shia Kapos' must-read rundown of political news in the Land of Lincoln May 31, 2024 View in browser   By Shia Kapos TGIF, Illinois . The 2024 presidential campaign just got a new talking point. TOP TALKER PLEDGING ALLEGIANCE: Illinois Republicans and a certain former gove...

Meta removes disinformation campaigns / Google Maps disables features in Ukraine / NIO seeks HK listing

Plus: OnePlus's upcoming phone can fully charge in 15 minutes. Inside.com Part of   Network February 28, 2022 Presented by Facebook parent company Meta said it took down two disinformation campaigns that targeted Ukraine and were spreading anti-Ukrainian propaganda. Russia-linked hacking group Ghostwriter, which perpetrated one of the campaigns, gained access to real Facebook accounts and posted YouTube videos "portraying Ukrainian troops as weak and surrendering to Russia," according to Meta. The company said it blocked phishing domains that Ghostwriter used to gain access to the accounts and beefed up security. More: Meta described the second disinformation campaign as "coordinated inauthentic behavior" made up of about 40 accounts, pages, and groups on Facebook and Instagram.  The campaign ran several websites disguised as independent...

Archive

Show more

Bottom Widget

Bottom Widget

font.css

videocontrol.css

GitFootCSS

BottomEdge Widget Title

Copyright © 2023 IdiotMail