Google accepted the resignation of prominent AI researcher Timnit Gebru, who's known for her work on algorithmic and facial recognition biases. Gebru, an advocate of diversity in tech and one of the few Black women in a senior AI role, says she was effectively fired over an email that she sent to colleagues. The email had expressed her frustration over a managerial request to retract a research paper, as well as the company's alleged treatment toward women and minority employees.

Background:

• Gebru is technical co-lead of Google’s Ethical Artificial Intelligence Team and a co-founder of Black in AI, a community of Black AI researchers.
• She's well known for her work with Algorithmic Justice League's Joy Buolamwini and others, which found higher disparities in facial recognition error rates between lighter-skinned males and darker females.

The controversy:

• Google recently asked Gebru to retract an AI ethics paper she co-wrote and submitted for submission at a conference. The company had wanted her to at least remove the names of Google employees from the paper, which examines ethical issues in language models.
• Gebru asked Google Research VP Megan Kacholia for an explanation. Without more discussions, Gebru said she would resign from the company.
• Subsequently, Gebru sent an email to company researchers, Google Brain Women and Allies, referring to the AI paper. She also made comments about how Google has failed to hire more women during the pandemic, among other concerns.

The fallout:

• The following day, Gebru said she received an email from Kacholia, reportedly approved by Google AI head Jeff Dean, saying that the company couldn't meet her demands and would accept her resignation.
• Google cited Gebru's email to brain group employees, saying it reflects "behavior that is inconsistent with the expectations of a Google manager."
• Gebru has since received support from fellow co-workers, AI experts, and others in the tech industry, including former Reddit CEO Ellen Pao, NY Times reporter Davey Alba, and Washington Post tech reporter Drew Harwell.
• "First and foremost I need an employment lawyer ASAP," Gebru tweeted.

Facebook introduced its ReBeL algorithm, a general gaming AI bot that it says excels at games like poker. ReBeL achieved superhuman performance in heads-up no-limit Texas Hold’em, according to the company, which called it "a major step toward creating ever more general AI algorithms."

More:

• ReBeL is the first AI to enable RL+Search, or combining reinforcement learning with search, in imperfect-information games, though it can work in all two-player zero-sum games.
• Facebook said ReBeL - which stands for Recursive Belief-based Learning - uses "far less domain knowledge" than previous poker AI bots.
• The algorithm relies on knowing a game's exact rules, like in poker and Go, though it can be "problematic in real-world interactions."
• Facebook's open-sourced its implementation for the game Liar’s Dice.

AI at small vs. large companies

MIT Technology Review Insights surveyed more than 300 business and tech leaders about their organizations' AI plans. Here's what they found:

• 62% of respondents said their organization is using AI technologies.
• 80% of larger organizations (more than $500M annual revenue) said they deployed AI, while small organizations (less than $5M in revenue) were slightly below average at 58%.

• Fewer than 5% of large organizations had no AI plans vs. 18% of smaller ones...

More details about Danish architect Bjarke Ingels' AI City in China were revealed during a global tech conference this week. The smart city, planned for Chongqing in southwest China, would rely largely on next-gen technologies and AI to gather and crunch data, which could help it better meet residents' needs.

More:

• According to Reuters, the latest plans raised some eyebrows as experts worry about mass surveillance, lack of privacy, and data sharing with governments.
• The city is being planned by Danish architecture firm BIG and Chinese tech company Terminus Group.
• Its first region, Cloud Valley, spans 33 acres with two adjacent plots that “mimic each other’s opposites,” similar to yin and yang. This phase has been under construction since April.
• Plans call for a protected network of courtyards with public functions, green rooftops for open-air events, a Robot Museum, and AI exhibition spaces, markets, and cafés. The project aims to incorporate self-driving cars, robotic vehicles, and e-bikes for its own "smart mobility" system.
• An AI virtual housekeeper named Titan would assist residents by presenting schedules and choosing their breakfast, for example.
• China says more than 500 campus-style "smart" cities are being built across the country.

No-code AI platform Signzy, which makes it possible for financial service companies to automate risk and compliance processes, raised $3M from Vertex Ventures. Altogether, the platform has raised $8.4M in total. With the money, the no-code platform aims to improve its AI platform and expand its global sales team.

More:

• Last month, the five-year-old Signzy raised $5.4M in a round led by Arkam Ventures and Mastercard.
• The news comes three weeks after Signzy today hired Arun Rajkumar as its new Vice President of Technology, where he is leading new tech research and development (R&D) efforts that will focus on AI and Computer Vision (CV) tech.
• At current, the Bangalore-based Signzy works with the four largest banks in India as well as 136 other banking and financial services worldwide.
• Check out this list of useful top no-code machine learning platforms.

A version of this story first appeared in Inside NoCode, our newsletter focused on no-code tools and non-traditional computer programming.

QUICK HITS: 

• During this holiday season, find out how SimpliSafe is protecting over 3 million homes with less markup and more security.*
• MIT machine learning models show that Moderna, Pfizer, and other COVID-19 trial vaccines may not cover people of Black or Asian ancestry as well as they do for white people.
• New non-profit MLCommons released its first effort: a public 86,000-hour speech dataset for AI researchers called the People’s Speech dataset.
• A hiring algorithm reflected biases against female job candidates using data from human reviewers, according to a new University of Melbourne study.
• Synthesia released a personalized video greeting with an AI Santa, who can say whatever users type in.
• Google bought data management company Actifio, terms undisclosed.
• Revtown’s changing the denim game by applying the best parts of workout clothes—comfort, flexibility & durability—to jeans.*

*This is a sponsored post.

Tweet of the Day: A $7.5M Microsoft donation will allow Code.org to offer AI/ML curriculum and more to students from elementary through high school.