The liar’s dividend

Hi, it’s Josh in New York. There’s no software program that can reverse years of damaged trust. But first...Three things you need to know to [View in browser]( [Bloomberg]( Hi, it’s Josh in New York. There’s no software program that can reverse years of damaged trust. But first... Three things you need to know today: • TikTok’s CEO will [meet with EU regulators]( • Apple warned [holiday sales could slow]( • A Boeing website was hit [by a cyberattack]( A ‘theoretical impossibility’ The Israel-Hamas war has highlighted how online deception can have an impact during periods of contentious public discussion. Some images purporting to show conditions in the region were [actually taken years ago](, while authentic photographs have been [dismissed as deepfakes]( after tools designed to identify AI-generated content have incorrectly flagged them. Deepfakes are one of the many AI-related issues that the Biden administration tries to tackle in its [executive order on artificial intelligence]( — the US’s most serious attempt yet to confront the challenges of AI. It [instructs the Commerce Department]( to study ways to identify and mark so-called synthetic content, as well as to authenticate content and track its true origin. It’d be great to have these things, but there are substantial challenges ahead, both technological and otherwise. There’s expert consensus that today’s systems for identifying content created by AI are gameable. Soheil Feizi, an associate professor of computer science at the University of Maryland whose research has demonstrated ways to break systems that detect AI [text]( and [imagery](, calls it a “theoretical impossibility” that future systems will be foolproof, either, since the basic AI models will always improve faster than the tools to detect what they produce. “There won’t be any reliable method to tackle these problems, even in the future,” he said.  There’s less agreement about what to do with these flawed tools. Feizi argues that imperfect watermarking could expose or deter less sophisticated attacks, even if it didn’t stop everything. The counterargument is that weak AI detection further breaks down trust each time it misses a deepfake or flags an authentic image as a phony and that it’s better not to pretend they work. But others have likened AI detection technology to locking the door to your house. Sure, someone could break the hinges with a crowbar or sneak in through a window, but it’s better than nothing. The more promising technology focuses on verifying where a piece of content comes from. In its most basic form, this would work by having the creator of the content “sign” it with a cryptographic code. Web browsers, social networks or other tools could automatically check this signature and signal to users that it’s legitimate. Similar systems are already used to authenticate the source of certain software programs and are difficult to break. Companies have sought to [develop standards]( for content authentication, but there’s some logic in having a government involved as well, said Hany Farid, a professor at the University of California, Berkeley, who studies digital forensics, human perception and misinformation. Farid said such a system could be paired with watermarking technologies to significantly clean things up online. “The most sophisticated, committed actors will find ways around this,” he said of AI detection technology. “But if we can lop off the bottom 75% of the problem, I’m going to declare victory.” However, technology is only part of the challenge. For years, people have been learning not to trust what they see online, especially things they don’t agree with. Concern about deepfakes — even if it’s warranted — could aggravate the damage already done by years of discussion about lower-tech misinformation campaigns on social media. Experts have begun to refer to this as the “[liar’s dividend](” because people looking to spread misinformation and distrust benefit as confusion and cynicism increases. Undoing the liar’s dividend will take more than good technology and an endorsement from governments and big companies. “When you don’t trust the institutions, this stuff doesn’t work,” said Farid. “It’s not just that you believe things that are factually incorrect. It’s that you believe the government, the media and the companies are keeping it from you.” In the long run, the only way to have a healthy internet in the AI era is to rebuild that trust. That’ll take a lot more than a presidential order. —[Joshua Brustein](mailto:jbrustein@bloomberg.net) The big story A Silicon Valley group backed by billionaires said it has enough land to [build a utopian city in Northern California](. The company behind the California Forever project said it wants to build a sustainable, walkable city that will create jobs. One to watch [Watch the Bloomberg Technology TV interview]( with Siemens CEO Roland Busch on his investment in US manufacturing. Get fully charged Elon Musk’s mysterious AI project will be released to a [select group of people](. OpenAI’s Dall-E 3 is so good that it’s stoking an artist [revolt against AI scraping](. Siemens will invest $510 million in the US and [build a new plant in Texas](. Wall Street excitement for Tesla’s stock [seems to be slowing](. More from Bloomberg Get Bloomberg Tech weeklies in your inbox: - [Cyber Bulletin]( for coverage of the shadow world of hackers and cyber-espionage - [Game On]( for reporting on the video game business - [Power On]( for Apple scoops, consumer tech news and more - [Screentime]( for a front-row seat to the collision of Hollywood and Silicon Valley - [Soundbite]( for reporting on podcasting, the music industry and audio trends - [Q&AI]( for answers to all your questions about AI Follow Us Like getting this newsletter? [Subscribe to Bloomberg.com]( for unlimited access to trusted, data-driven journalism and subscriber-only insights. Want to sponsor this newsletter? [Get in touch here](. You received this message because you are subscribed to Bloomberg's Tech Daily newsletter. If a friend forwarded you this message, [sign up here]( to get it in your inbox. [Unsubscribe]( [Bloomberg.com]( [Contact Us]( Bloomberg L.P. 731 Lexington Avenue, New York, NY 10022 [Ads Powered By Liveintent]( [Ad Choices](