|
|
|
|
|
AT&T Reuses AI Building Blocks; Schools Monitor Returning Students
|
|
|
|
|
|
Welcome back. Creating commercial artificial-intelligence capabilities is a time-consuming effort, especially when your software developers keep reinventing the wheel. All too often, AI experts in different parts of a company forge the same basic AI building blocks each time they develop new tools, unnecessarily duplicating each others’ work. To avoid such digital redundancies, some companies are launching repositories of software code, data components and AI models that can be reused again and again.
Known as "feature stores," the resources are typically accessible to a company’s data scientists via a portal-like marketplace or searchable catalog. Uber Technologies Inc., a pioneer of the concept, has used AI models for its ride-sharing apps to develop its Uber Eats delivery service, said Smitha Shyam, an Uber director of engineering who leads the company’s AI efforts.
AT&T Inc., the latest company to launch a feature store, said its data scientists are leveraging reusable components to build new fraud-detection apps, among other tools.
|
|
|
|
|
|
AT&T’s 'feature store' opened this month and will be initially available to the team responsible for the bulk of the company’s data science work.
PHOTO: LARRY W SMITH/SHUTTERSTOCK
|
|
|
|
Digital Convenience Stores
|
|
AT&T Inc. is the latest company to open an in-house ‘feature store’ to act as a digital tool shed of reusable data components, a move companies hope will provide their software developers ready-made building blocks for new AI tools, rather than starting from scratch, the WSJ’s John McCormick reports.
What’s inside. AI feature stores are typically stocked with the complex data elements that make up an AI model, along with the description of what the feature is, what it does and how it was built.
Who gets in. AT&T says its store will be open initially to Chief Data Officer Andy Markus’s data science team, allowing access to data science teams across its business units before the end of the year.
More locations. Feature stores emerged several years ago with big tech companies and advanced business users of AI, and are being embraced by more companies as they compile the data and talent needed to push AI development.
|
|
|
Researchers are developing innovation-prediction algorithms in a bid to take a quantitative approach to examining how quickly technologies improve, the WSJ's Christopher Mims reports.
|
|
|
PHOTO: GETTY IMAGES/ISTOCKPHOTO
|
|
|
How it works. Using more than 40 years of patent data, combined with a custom empirical dataset on the rates of technological improvements, MIT researchers trained a prediction algorithm to determine the pace of progress for patented technologies.
|
|
|
Who benefits. By forecasting the advent of new technologies, investors, company leaders and government planners can make better-informed decisions about where to direct money, time and attention.
The upshot. Viewed across decades, the research finds, individual technologies change at a surprisingly steady rate based on the underlying physics that any given technology is built on, rather than the brilliance of inventors.
|
|
|
|
|
“We’re optimizing how the scientists spend their time; we’re not rebuilding features each time.”
|
— AT&T Inc. Chief Data Officer Andy Markus on the company’s new AI feature store
|
|
|
|
|
|
|
|
|
ILLUSTRATION: BENEDETTO CRISTOFANI
|
|
|
|
|
As students return to real-world classes, some school districts are using AI-enabled software to scan their communications and web searches on school-issued devices, or in some cases devices that are logged in via school networks, for signs of suicidal thoughts, violence against fellow students, bullying and more, the WSJ's Julie Jargon reports.
How it works. When the trained AI algorithm recognizes certain key phrases, these systems typically send an alert to school officials who determine if an intervention with students and their parents is warranted.
Supporters. School administrators say such surveillance is more important than ever as students return to the classroom after 18 months of pandemic-related stress, uncertainty and loss.
Critics. Others say the use of smart software to track students raises questions about privacy, misuse and students’ ability to express feelings freely or search for answers.
Apple a day. Separately, The Wall Street Journal reports that Apple Inc. is working on technology to help diagnose depression and cognitive decline, expanding the scope of its burgeoning health portfolio.
|
|
|
|
|
Goldman Sachs is acquiring home-improvement lender GreenSky.
PHOTO: BRENDAN MCDERMID/REUTERS
|
|
|
|
Lending Apps Seek Bankification
|
|
Several notable digital lending companies that began with some form of partnership, wholesale funding or marketplace model, including AI-powered lending platforms and other nonbank lenders, are in some stage of shifting into a banking model, The Wall Street Journal reports.
Take it to the bank. Home-improvement lender GreenSky recently agreed to be acquired by Goldman Sachs, while LendingClub became a bank after buying Radius Bancorp, and Square acquired a banking license to issue business loans.
Happy investors. Shares of Upstart Holdings, an AI lending platform that works with partner banks and investors, are up sevenfold this year amid a surge in loan volume, while banks in the S&P Composite 1500 index are up about 25% on the year.
|
|
|
|
81%
|
The share of third- to 10th-grade teachers who said their school uses some form of student-monitoring software, including AI-enabled tools, according to a Center for Democracy and Technology survey
|
|
|
|
|
|
|
Wells Fargo plans to hire more cloud security engineers as a result of its digital infrastructure initiative.
PHOTO: DAVID PAUL MORRIS/BLOOMBERG NEWS
|
|
|
|
Wells Fargo shifts to public cloud. In a bid to boost its AI chops, among other initiatives, Wells Fargo & Co. has unveiled a decadelong initiative to update its digital infrastructure in part by moving its workloads to cloud services from Microsoft Corp. and Alphabet Inc.’s Google. (The Wall Street Journal)
Fintech startup Pagaya to go public. New York and Tel Aviv-based Pagaya Technologies Ltd., which operates an AI network designed to make bank transactions easier, is going public through a merger with a special-purpose acquisition company that values the financial-technology startup at about $9 billion. (The Wall Street Journal)
Indo-Pacific pact includes AI. A security partnership struck last week by the U.S., the U.K. and Australia that focused on the development of nuclear submarines includes cooperation on research in areas like AI, cybersecurity and quantum computing. (The Wall Street Journal)
Realtors dig into data. More real-estate firms are turning to AI to plow through millions of documents for data on property values, debt levels, home renovations and homeowners, in a bid to get an edge on buying, selling and home financing. (CNBC)
Bots just want to help. The latest installment of a multiyear study by Stanford University says AI is shaping up to be a tool best used to augment human activity, not automate it. (ZDNet)
Robot-written code includes bugs. Some AI-powered tools designed to help developers write software code are re-creating many of the same human errors that lead to bugs and other glitches. (Wired)
GPT-3 rife with bias, study shows. Stanford researchers say language tests on AI-powered automated text generator GPT-3 reveal a disproportionate association with violence when using the term Muslim, compared to similar tests using the term Christian. (Vox)
Minesweepers harness drones. Engineers are turning to AI-equipped drones to find and disarm long-forgotten landmines, hoping to replace the risky task, often carried out by nonprofit humanitarian organizations, of sweeping suspect sites on foot with magnetometers or other handheld instruments. (Phys.org)
Afro-Latina bot unveiled. Tech advocacy group Create Lab Ventures has created what it calls the first bilingual AI-powered Afro-Latina bot, aiming to promote tech and media skills in underserved schools. (Black Enterprise)
|
|
|
Alphabet Inc.'s Google is buying a Manhattan office building for $2.1 billion, one of the clearest signals yet of Big Tech’s growing appetite for office space, even as these companies embrace remote work. (The Wall Street Journal)
Twitter Inc. said it agreed to pay more than $800 million to settle a consolidated class-action securities lawsuit alleging the social-media company deliberately misled investors about user engagement in 2015. (The Wall Street Journal)
European cloud-services provider OVHcloud plans to launch a possible initial public offering, the company said, a deal that could value the business at more than $4.7 billion. (The Wall Street Journal)
ByteDance Ltd., the maker of the hit short-video mobile app TikTok, said that it would restrict access to Douyin, the Chinese version of the app, to 40 minutes a day for users under 14 years old. (The Wall Street Journal)
|
|
|
|
|
|