Maps are the next frontier for deepfakes

PLUS: The latest stories on how we move, work, and live.

View this email in your browser  |   Manage newsletter subscriptions 06.01.21 Welcome to the new Fast Forward! We'll now be coming to you every Monday and Thursday with what's next in artificial intelligence and technology. In today's edition, we peer into a future in which automatic features on cars are so good people spend more time behind the wheel—but maps can't be trusted. Still, some things never change: Politicians can't seem to get AI regulation right.     Deepfake Maps Could Really Mess With Your Sense of the World In a new paper, University of Washington professor Bo Zhao employed AI techniques similar to those used to create so-called deepfakes to alter satellite images of several cities, Will Knight reports. Zhao used an algorithm called CycleGAN, which trains an artificial neural network to recognize the key characteristics of images. A second algorithm refines the performance by trying to detect when an image has been manipulated. Such altered maps could mislead governments, or sow misinformation and doubt on social media. Why It's Scary As the technology advances, sophisticated "deepfake" satellite images could be used to hide weapons installations or wrongly justify military action. When Driving Is (Partially) Automated, People Drive More Self-driving cars aren't here yet, but as Aarian Marshall reports, partially automated car features, such as Tesla's Autopilot, are already changing driving patterns. In a survey of 630 Tesla owners, researchers found that motorists with partial automation drive on average 4,888 more miles per year than similar owners without the feature. In particular, people are opting for longer commutes and going farther on weekend road trips. Drivers with partial automation feel "more relaxed, less tired, less stressed," UC Davis researcher Scott Hardman says. Where This Is Heading If drivers log more miles as automated features become more common, it could mean more greenhouse gas emissions, more strain on roads, and fewer riders on public transportation. Dumbed Down AI Rhetoric Harms Everyone The European Commission's regulatory proposal on AI mostly just underscores democracies' confusing rhetoric on AI, argues Justin Sherman. A meaningful progress begins with resolving the discourse's many contradictions and unsubtle characterizations. A lengthy discussion of the risks of AI harming marginalized groups doesn't mesh with giving law enforcement broad authority to use facial recognition in practice. Tech groups, including Silicon Valley firms, have pushed back on the proposed regulations, leaning on vague and contradictory explanations. Framing the competition as an AI "arms race" inaccurately suggests all AI development, particularly between the US and China, is winner-take-all. The Bottom Line More AI regulation is necessary, but policymakers should avoid perpetuating the false dichotomies of privacy vs. security and regulation vs. competitiveness.     Need to Know   Eric Ravenscraft assesses Google's AI phone call features. Should you let your Android make reservations, ask unknown callers to identify themselves, and wait on hold for you? Khari Johnson looks into Microsoft and OpenAI's plan to use GPT-3, an advanced model for generating text, to write computer code based on natural language descriptions. Georgetown disinformation experts demonstrated how effectively the text-generation algorithm GPT-3 can write tweets that amplify misinformation or push a faulty narrative, Will Knight writes. Gilad Edelman tries out CaliberAI's new tool that's like spell-check for libel. It's meant for newsrooms now, but could be used to moderate social media posts. A study of 10,000 images found that Twitter's automatic cropping algorithm favored white faces and women, Khari Johnson writes.       So, This Happened Facebook and Twitter both blamed their automated content moderation systems for blocking the accounts of Palestinian activists during the recent Israeli-Palestinian violence. (The Washington Post) A "poorly worded tweet" forced insurance start-up Lemonade to clarify that it doesn't use emotion recognition and doesn't use AI to automatically reject claims. (Lemonade) Utah's DOT said it will begin testing Payver, which uses crowd-sourced video and machine learning to determine when roads need to be repaired. (TechCrunch) The Royal Navy unveiled a nuclear-powered submarine with movements and actions completely controlled by AI. (BBC)     Until Next Time   Why are cicadas SO LOUD? We had an entomologist break it down.     More Essential Reads How to Master Google's AI Phone Call Features BY ERIC RAVENSCRAFT Your Android phone can do a lot of fancy tricks, from making reservations with Duplex or screening calls. Here's how to get the most out of them.   5-minute read AI Could Soon Write Code Based on Ordinary Language BY KHARI JOHNSON Microsoft reveals plans to bring GPT-3, best known for generating text, to programming. "The code writes itself," CEO Satya Nadella says.   3-minute read AI Can Write Disinformation Now—and Dupe Human Readers BY WILL KNIGHT Georgetown researchers used text generator GPT-3 to write misleading tweets about climate change and foreign affairs. People found the posts persuasive.   3-minute read These Ex-Journalists Are Using AI to Catch Online Defamation BY GILAD EDELMAN CaliberAI wants to help overstretched newsrooms with a tool that's like spell-check for libel. But its potential uses go far beyond traditional media.   7-minute read GET WIRED Memorial Day Sale! Ending Soon: Get WIRED for $10 $5. You can't see into the future, but you can subscribe to WIRED. Subscribe now.             This e-mail was sent to you by WIRED. To ensure delivery to your inbox (not bulk or junk folders), please add our e-mail address, wired@newsletters.wired.com, to your address book.   View our Privacy Policy Unsubscribe or manage your newsletter subscriptions   Copyright © Condé Nast 2021. One World Trade Center, New York, NY 10007. All rights reserved.