๐ NYC's AI Scanner False Alarm, Google Enhances Chrome Security, AI Ethics Loophole Discovered
Todayโs pick
NYC Has Tried AI Weapons Scanners Before. The Result: Tons of False Positives. Last week, Mayor Eric Adams announced that he was kicking off a 90-day waiting period to test weapons scanners on New York City's subway system. The scanners that he demonstrated, provided by Evolv Technology, purport to mix physical detection technology with artificial intelligence to indicate when someone is carrying a gun or a knife, and have been used in City hospitals before, something Adams alluded to at the press conference at Fulton Street station. By Felipe De La Hoz via Hell Gate
๐: NYC's "AI" "weapon scanners" produce 85% false positives!!1!! AKA they don't detect weapons-- they give cops a pretext to harass ppl while lining the pockets of vendors. In the meantime, NYC is shutting libraries on weekends citing budget constraints. - Meredith Whittaker (@mer__edith)
Google to Fight Cookie Hijacking With Encryption Keys for Chrome Browser. 'Device Bound Session Credentials' will tap the TPM chip in today's PCs. Google is testing it with some users running the beta version of Chrome. Stealing a password isn't the only way a hacker can break into your online accounts. Itโs long been known that malware can also loot a browserโs cookies to hijack your login sessions. Now Google is trying to thwart the threat with a new prototype feature for the Chrome browser. By Michael Kan via PCMag
๐: Fighting cookie theft using device bound sessions - Gianluca Varisco (@gvarisco)
The best way to reach new readers is through word of mouth. If you click THIS LINK in your inbox, itโll create an easy-to-send pre-written email you can just fire off to some friends.
Anthropic researchers wear down AI ethics with repeated questions. How do you get an AI to answer a question itโs not supposed to? There are many such โjailbreakโ techniques, and Anthropic researchers just found a new one, in which a large language model (LLM) can be convinced to tell you how to build a bomb if you prime it with a few dozen less-harmful questions first. They call the approach โmany-shot jailbreakingโ and have both written a paper about it and also informed their peers in the AI community about it so it can be mitigated. By Devin Coldewey via TechCrunch
๐: My biggest takeaway re: many-shot jailbreak is actually the unreasonable effectiveness of this Cautionary Warning Defense prompt: jailbreak effectiveness tanks from 61% to 2%. I've found final-message appended reminders useful for boosting instruction following generally too - Paul Calcraft (@paul_cal)
Welcome to the AI gadget era. I'm just going to call it: we'll look back on April 2024 as the beginning of a new technological era. That sounds grandiose, I know, but in the next few weeks, a whole new generation of gadgets is poised to hit the market. Humane will launch itsย its voice-controlled AI Pin. Rabbitโs AI-powered R1 will start to ship. Brilliant Labsโ AI-enabled smart glasses are coming out. And Meta is rolling out a new feature to its smart glasses that allow Metaโs AI to see and help you navigate the real world. By David Pierce via The Verge
Microsoft, Quantinuum claim breakthrough in quantum computing. Microsoft (MSFT.O), opens new tab and Quantinuum on Wednesday said they have achieved a key step in making quantum computers a commercial reality by making them more reliable. The move is the latest in a race to perfect quantum computing in which tech firms such as Microsoft, Alphabet's Google (GOOGL.O), opens new tab and IBM (IBM.N), opens new tab are jostling with both rivals and nation states to create machines that take advantage of quantum mechanics to promise speeds far faster than conventional silicon-based computers. By Stephen Nellis via Reuters
Thanks for reading to the bottom and soaking in our Newslit Daily fueled with highlights for your morning.
I hope you found it interesting and, needless to say, if you have any questions or feedback let me know by hitting reply.
Take care and see you tomorrow!
P.S. Want to advertise with us? Weโd love to hear from you.