|
|
|
|
Banks Chase Fintech Upstarts for Smarter Services; More Talk Generated by GPT-3
|
|
|
|
|
|
|
|
Welcome back. The banking sector is playing catch up with technology firms in the race to develop and adopt smart software. AI-enabled tools promise to help lenders personalize services, evaluate credit risk and more. Elsewhere, researchers are dialing back some of the loftier promises of more advanced capabilities, like OpenAI’s language-generating system GPT-3, highlighting shortcomings in comprehension and logic.
|
|
|
|
|
|
|
|
KeyBank’s head of data science said she thinks AI innovation will become more concentrated at financial-technology companies. PHOTO: TY WRIGHT/BLOOMBERG NEWS
|
|
|
|
Facing less red tape than banks, large technology firms are carving out a niche in financial services by leveraging their AI know-how to build advanced lending and risk-management tools, WSJ’s Jared Council reports.
Big spenders. Google plans to introduce checking accounts next year, a move Amazon is eyeing in talks with financial institutions. Apple last summer rolled out its own credit card.
High interest. Big tech firms see financial services as a way to glean valuable data from users, turning out a range of fintech tools via existing cloud-based AI systems without incurring major capital costs.
|
|
|
|
|
|
|
LinkedIn is rolling out software designed to flag bias in artificial intelligence algorithms, Jared reports. The professional networking site said it is using the tool kit to review the thousands of algorithms used to run its platform. LinkedIn also said it has made the technology available to other companies.
How it works. The LinkedIn Fairness Toolkit, or LiFT, can detect whether an AI algorithm generates results that over- or underrepresent groups of people by gender, race, age or other attribute, LinkedIn says.
Why now? The site is unveiling the tool amid a national reckoning on racial injustice, which has heightened attention on AI algorithms that could perpetuate bias.
|
|
|
|
|
|
|
$165 million
|
|
The amount of cash data-analytics company Palantir Technologies burned through last year. The 16-year-old venture disclosed its losses ahead of a long-delayed public listing.
|
|
|
|
|
|
|
Language-Generating Systems
|
|
|
|
Can We Talk?
Hype around AI is nothing new. But since OpenAI unveiled GPT-3, an AI-powered language generator, the tech world has been especially abuzz over its ability to produce coherently written responses to queries. Now some of the system’s early testers are tamping down so-called GPT-3 shock by pointing out its limitations—not least cloudy logic.
Start making sense. Even when GPT-3 is primed with enough material with which to generalize, it still needs a human in the loop to separate out oddities, David A. Price writes in WSJ, citing its interpretation of the famous opening line of Jane Austen’s “Pride and Prejudice” as “men are very vain and want to be seen as wealthy.”
|
|
|
Reasonable doubt. Despite being grammatically correct, the system’s “comprehension of the world is often seriously off, which means you can never really trust what it says,” industry insiders Gary Marcus and Ernest Davis write in the MIT Technology Review.
|
|
|
|
|
|
“Remove the door. You have a table saw, so you cut the door in half and remove the top half.”
|
|
— OpenAI’s language-generating system GPT-3 responding to a query by researchers about how to fit an oversized table into a living room.
|
|
|
|
|
|
|
|
Grain bins crumpled by severe windstorms in Luther, Iowa. The Energy Department and Microsoft aim to come up with 10 to 30 AI-powered tools to help emergency workers in the early stages of an extreme natural event. PHOTO: DANIEL ACKER/GETTY IMAGES
|
|
|
|
Algorithms to the rescue. The U.S. Department of Energy and Microsoft announced a partnership to develop AI tools aimed at helping first-responders react to fast-changing natural events, such as floods and wildfires. (WSJ)
Can robots invent robots? A scientist in Missouri has sued the U.S. Patent and Trademark Office for rejecting a pair of patent applications that named an AI system as the inventor of a beverage container based on fractal geometry and a light beacon designed to attract more attention. (Reuters)
Waterdrop raises more capital. The Chinese online insurance-technology platform has raised an additional $230 million in a new funding round led by Swiss Re and Tencent Holdings. The $2 billion venture says the cash will go into tapping AI to accelerate medical and health-care projects. (Reuters)
Facebook sound simulator. The social-media platform is developing a system that uses simulated indoor sounds to train AI to locate objects in three-dimensional spaces, such as a glass on a table or a misplaced smartphone. Facebook also unveiled a mapping tool designed to help AI systems better understand indoor spaces. (CNN)
Maverick! AI-powered aviation software beat a human F-16 pilot in a simulated dogfight, which included five different basic fighter maneuver scenarios overseen by U.S. Air Force officials. The unnamed human pilot is a recent graduate of the Air Force Weapons School’s F-16 Weapons Instructor Course with more than 2,000 hours of airtime. (Air Force Magazine)
City officials scrap smart tools. Local government district councils in the UK are discontinuing the use of smart software in making decisions about welfare benefits and other issues, amid criticism over a lack of transparency in deploying AI in public-services administration. (The Guardian)
|
|
|
|
|
|
|
Chinese-owned TikTok sued the U.S. government in federal court, challenging President Trump’s executive order that would effectively ban the video-sharing app if it doesn’t find an American buyer for its U.S. operations. (WSJ)
An hours-long outage on videoconferencing app Zoom Monday prevented businesses, schools and other users from starting or joining remote meetings and webinars. (WSJ)
Microsoft in a court filing said payment restrictions by Apple on 'Fortnite' videogame maker Epic Games could hamper its own game-development business. (WSJ)
|
|
|
|
|
|
