Google is restructuring its responsible AI efforts after the effective firing of AI researcher Timnit Gebru. In a blog post, Google said engineering VP Marian Croak will now oversee "a new center of expertise" at Google Research that's focused on the responsible development of AI technologies.
More:
- According to Bloomberg, Croak’s appointment is meant to centralize and stabilize Google's AI efforts after Gebru's December departure, which has stirred divisions at the tech giant. Croak, a Black Google executive, will report directly to Jeff Dean, the lead of Google AI; she will also oversee Google's engineering fairness products.
- After Gebru’s exit, Google's AI team published a letter criticizing the company for her dismissal and calling for manager Megan Kacholia to be replaced. Kacholia led the company's ethical AI researchers, of which Gebru was a member.
- The ethical AI team includes people working on computer vision, natural language processing, and other AI disciplines. The restructuring effectively removes Kacholia from this role.
- In a new YouTube video, Croak acknowledged the “dissension” in Google's ethical AI efforts, which were exacerbated when Gebru, a prominent Black researcher, was asked to take her name off an AI research paper that criticized technologies made by Google and others.
- Croak said she'd like to focus on more "diplomatic" conversations concerning AI responsibility and fairness, "so we can truly advance this field.”
- Some Google AI team members, including Dr. Alex Hanna, have criticized Google for not telling them about the changes sooner. In a series of tweets, Hanna called it "nothing short of a betrayal."

BLOOMBERG
|
|
Technology being developed to read and alter brain activity could pose privacy risks, warn ethicists, scientists, and Science News readers. According to a survey of readers, privacy was the most worrying aspect of neurotechnology being developed by Elon Musk's Neuralink and other companies, beating out autonomy and fairness by a wide margin.
More:
- Musk founded Neuralink, his neural tech company, in 2016 to develop devices that could eventually sync the human brain with AI.
- Neuralink unveiled last year a pig called Gertrude that had a computer chip in its brain that served as a brain-to-machine interface. In late Janaury, Musk said Neuralink has implanted one of its brain-chips into a monkey, allowing the animal to play video games with his mind.
- Rafael Yuste, a neurobiologist at Columbia University, told Science News that scientists are getting close to being able to pull private information from people's brains, such as what someone is looking at or hearing.
- Yuste thinks that even subconscious thoughts might be discoverable using technology: "That is the ultimate privacy fear, because what else is left?" he asked.
A version of this story first appeared in Privacy XFN, our newsletter curating the best reads at the intersection of data privacy and tech.
SCIENCE NEWS
|
|
Strong adoption of AI tools helped drive higher-than-expected sales in Baidu's last quarter, according to Reuters. The Chinese tech giant is seeking to diversify by providing AI technologies for "mature use cases."
More:
- Yesterday, Baidu reported quarterly revenue above analysts' expectations. Revenue rose 5% to 30.26B yuan ($4.69B), above estimates of 30.06B yuan.
- After the earnings report, Baidu founder Robin Li said Baidu is now "a leading AI ecology company," thanks to its long-term focus on tech research. In recent years, it has invested in areas such as autonomous driving, language learning, and voice interaction, which reports say are finally paying off.
- The company — which is already involved in the AI semiconductor space and sells an AI-powered platform-as-a-service in China — continues to diversify. Baidu said it's seeking to raise money for a standalone AI chip company. It's currently in talks with VC firms to invest in the business, which would manufacture AI chips for automakers and other industries.
- In 2019, Baidu and Samsung partnered to mass-produce Kunlun, a cloud-to-edge AI chip that can process a large amount of data for AI applications.
REUTERS
|
|
An international team of researchers is working on an "electronic nose" that relies on dogs and machine learning to interpret various scents. In a new study, the researchers describe how they used the olfactory senses of dogs to help develop a machine learning algorithm, which could one day detect cancers as accurately as canines.
More:
- Their proof-of-concept method focused on canine olfaction and machine odor analysis to detect prostate cancer. Their algorithm was trained on 50 urine samples from cancer patients and a control group.
- The machine learning system achieved a more than 70% success rate in follow-up testing, similar to the dogs' results.
- Andreas Mershin, a research scientist at the MIT Center for Bits and Atoms, said the research confirms that they can train an AI system to mimic the accuracy of dogs. Due to the small sample size, the research would require a larger experiment for validation.
- Their goal is to progress the technology further, possibly to create a small "scent detector" that could be incorporated into smartphones. Before that, the researchers would need to use more samples to raise the dogs’ cancer-detecting accuracy and train the artificial neural network "to match this performance."
THE SCIENTIST
|
|
Goldman Sachs is launching Marcus Invest — a Robo-advisory wealth management platform — as a further push to grow its consumer banking unit. Robo-advisory services are automated investment platforms that give investment advice without human intervention.
More:
- Marcus Invest, the Robo-advisory platform, will allocate and rebalance investments across a mix of ETFs, stocks, and bonds based on customers' risk tolerance and investment timeline. The offering is a bet against the recent retail investment frenzy as the platform doesn’t allow investors to trade single stocks.
- The platform will enable people to invest with a minimum of $1,000 at a 0.35% annual fee. It will be introduced by the second half of this year.
- The Robo-advisory strategy, which automatically builds and rebalances the diversified portfolio of funds, is already offered by a few firms. It is expected to manage $460B by 2022, up from $98.5B in 2016.
- According to Forbes, a number of enterprises are using AI in robot advisory. For example, Betterment's AI can "reduce taxes on transactions where machine learning algorithms select the specific tax consequences of the transactions." AI also can automatically allocate assets and determine which investments will result in the lowest taxes.
A version of this story first appeared in Inside Business. Read the full issue here.
|
|
Zillow's 3D Home app has a new machine-learning-powered feature that can generate interactive floor plans, showing room dimensions, and vantage points. Users can view still photos and AI floor plans in a shared interface, giving them a better feel for a home.
More:
- The feature, part of Zillow's virtual tour experience, generates the plans based on uploaded photos of the home.
- The idea is to "break down the barriers" between listing data and media, according to Josh Weisberg, Zillow’s VP of media experience. It helps home shoppers better visualize the relationship between photos and a house's layout, giving them a better sense of a home's features and space, he said.
- The AI system predicts and adds square footage and room dimensions to the virtual tours. Once finished, the interactive floor plan is automatically added to the Zillow listing and can be uploaded to multiple listing services.
THE VERGE
|
|
QUICK HITS:
- Thinking about investing in real estate? Distressed real estate markets provide tons of opportunity for savvy investors. Learn about investing in Caliber.*
- DARPA is working on an AI tool that's designed to curb "friendly fire" during joint U.S. military operations. Live tests are planned for 2024.
- German automaker Daimler has chosen Amazon Web Services to support its upcoming tests of self-driving trucks. Daimler's subsidiary Torc Robotics is aiming to achieve Level 4 self-driving technology, which means a vehicle can operate without human intervention.
- AEye, a developer of LiDAR technology for autonomous driving, will merge with CF Finance Acquisition Corp. III at a $2B valuation.
- States Title, a provider of AI-based solutions for real estate, raised $150M in debt financing from HSCM Bermuda.
- Boast.ai, producer of automation software to apply for R&D tax credits, secured a $100M credit facility from Brevet Capital.
- Why shouldn’t all investors have access to the best investing opportunities? Enter Titan, an elite investment manager for everyone.*
*This is a sponsored post.
|
|
|
|
|
Beth is a tech writer and former investigative reporter for The Arizona Republic. A graduate of the Walter Cronkite School of Journalism, she won a First Amendment Award and a Pulitzer Prize nomination for reporting on the rising costs of public pensions.
|
|
Editor
|
Charlotte Hayes-Clemens is an editor and writer based in Vancouver. She has dabbled in both the fiction and non-fiction world, having worked at HarperCollins Publishers and more recently as a writing coach for new and self-published authors. Proper semi-colon usage is her hill to die on.
|
|