|
DeepMind Pursues Real-World Autonomy; Pinterest Pins Growth on Neural Networks
|
|
|
|
|
|
Welcome back. Researchers working on advanced software that holds the promise of giving full autonomy to robots, among other applications, say they can’t make headway without having full autonomy themselves. Their corporate parents, wary of catching the attention of regulators, say they have good reason to keep a close eye—and a short leash—on units developing controversial tools like deep learning or facial recognition. Striking the right balance between caution and independence is at the heart of mounting tensions inside Google, which recently cut off talks with DeepMind, its deep-learning powerhouse.
|
|
|
|
|
Google bought U.K.-based DeepMind in 2014.
PHOTO: TOBY MELVILLE/REUTERS
|
|
|
|
Alphabet Inc.’s Google last month derailed a yearslong push by senior managers at DeepMind, its AI unit, to form an independent legal structure to pursue the sensitive research it does, WSJ’s Parmi Olson reports.
Mind over matter. Founded in 2010 and bought by Google in 2014, DeepMind specializes in building advanced deep-learning systems, a data-heavy technology that has caught the attention of regulators such as the European Union.
Ethical quandary. Google has dealt with controversies involving its top AI researchers and executives in the past, and this month unveiled plans to double the size of its ethics team.
|
|
|
|
New research finds that users are turned off by AI-enabled chatbots that project arrogance or overconfidence, but they’re also less likely to trust bots that seem incompetent, creating a tricky balance for software developers, The Wall Street Journal reports.
|
|
Artoo or Threepio? Ranjay Krishna, a Ph.D. candidate at Stanford University and one of the study's researchers, described the line between likeable and brainy chatbots as a “double-edged sword.”
|
|
|
Know-it-alls. The findings are based on feedback from participants who interacted with chatbots programmed to emulate toddlers, middle schoolers, recent graduates and trained professionals, among other demographics.
|
|
|
“Guided by safety and ethics, this invention could help society find answers to some of the world’s most pressing and fundamental scientific challenges.”
|
— DeepMind statement on benefits of deep learning.
|
|
|
|
|
|
|
Neural networks like Pinterest’s have artificial neurons that mimic the human brain and make connections such as bathroom-tile shoppers being interested in vanities.
PHOTO: WSJ|, PHOTOS: ISTOCKPHOTO/GETTY IMAGES
|
|
|
Pinterest AI Drives Growth
|
|
By leveraging neural-network technology, Pinterest Inc. is improving its success rate in finding the right eye-catching images to draw in users and boost ad-revenue numbers, WSJ’s John McCormick reports.
How successful? Neural networks are driving “nearly 100%” of growth, says Jeremy King, Pinterest’s senior vice president of engineering and a former executive vice president and chief technology officer at Walmart Inc.
Big bucks. By using AI to sort through unstructured data and tag users’ interests, the image-search site has more than doubled ad sales since 2018, the company says.
|
|
|
|
Algorithms Fill the Shelves
|
|
The number of books written about AI, machine learning and neural networks is expanding nearly as fast as the technology itself. Here are some of the best, as compiled by The Wall Street Journal.
|
|
Genius Makers (Dutton, 370 pages, $28). New York Times technology reporter Cade Metz tells the compelling story of the scientists who developed deep learning, a small group of researchers “who nurtured an idea for decades, often in the face of unfettered skepticism before it suddenly came of age.” Mr. Metz also notes that there is a long tradition of tech leaders promising “lifelike technology that was nowhere close to actually working.”
|
|
|
A Brief History of Artificial Intelligence (Flatiron, 262 pages, $27.99). British researcher Michael Wooldridge says he aims to temper the hyperbole around what AI can do, dismissing the notion of a computer with a full range of human intellectual capabilities as “nothing more than speculation.”
The Myth of Artificial Intelligence (Harvard/Belknap, 312 pages, $29.95). Computer scientist and entrepreneur Erik Larson argues that the fundamental error we make when thinking about AI is failing to recognize that “human and machine intelligence are radically different,” noting that general intelligence is not “an algorithm running in our heads.”
|
|
|
|
480 Million
|
The number of users drawn to Pinterest in the first quarter this year by using neural networks to predict their interests and target ads, the company says.
|
|
|
|
|
PHOTO: RICHARD DREW/ASSOCIATED PRESS
|
|
|
Ex-Goldman investor targets data. Sumit Rajpal, former co-head of Goldman Sachs Group Inc., has recruited experts in AI and data analytics, and plans to launch a private-equity firm focused on companies with troves of data. (The Wall Street Journal)
MIT Grad launches video manager. Netra, co-founded by MIT alumnus Shashi Kant, uses AI to identify activities, objects, emotions, locations and more in videos posted online, aiming to help YouTube, TikTok and other companies organize giant stores of videos. (SciTechDaily)
SoftBank backs AI tutor startup. Riiid Inc., a South Korean startup that uses a AI to create automated student tutorials, raised $175 million from SoftBank Group Corp.’s Vision Fund 2, funding it says will go to developing better standardized testing for public schools. (Reuters)
British Navy launches smart sub. The U.K. Ministry of Defence unveiled an fully autonomous nine-ton submarine built by MSubs, a Plymouth-based specialist in autonomous underwater vehicles, that is designed to operate up to 3,000 miles from the nearest naval base entirely on AI. (BBC)
China rover roams Mars. China’s space administration on Saturday said its solar-powered rover successfully touched down on Martian soil, the second autonomous vehicle to land and operate on Mars, after the U.S. rover earlier this year. (Associated Press)
Intelligent toilet tester. Duke University researchers are developing an AI-powered device designed for standard toilets that can analyze patients’ stool and provide data for the treatment of chronic issues, such as inflammatory bowel disease and irritable bowel syndrome. (Duke Today)
Connected car startup to go public. Wejo, a British firm that analyzes data from connected cars, is in talks to go public through a reverse merger with blank-check company Virtuoso Acquisition Corp., a move that would value the company at more than $1 billion (Reuters)
|
|
|
Germany’s competition regulator is investigating whether Google is dominant enough to be subject to the country’s new digital-competition law, broadening its efforts to brandish new powers against companies including Amazon.com Inc. and Facebook Inc. (The Wall Street Journal)
A chill has fallen over China’s new generation of tech giants, with stock in former market darlings such as Meituan, Pinduoduo Inc. and Kuaishou Technology dropping by more than one-third from highs reached earlier this year. (The Wall Street Journal)
Hangzhou Hikvision Digital Technology Co., the world’s largest maker of surveillance equipment, has long-established links to China’s military, including conducting a study with Chinese weapons experts and supplying cameras and drones to the country’s air force, according to a report by a surveillance-industry research company. (The Wall Street Journal)
|
|
|
|