Security researchers find evidence hackers are using ChatGPT and Claude to scan websites for vulnerabilities. In an election year, detecting AI-driven attacks on campaign websites and other infrastructure may be harder than spotting deep fakes. isc.sans.edu/diary/OpenAI... 🐘♻️
OpenAI Scans for Honeypots. Artificially Malicious? Action Abuse?, Author: Johannes Ullrich
Whenever AI feels intelligent, it's critical to remember that the Turing Test assesses the strength of an *illusion*. Some illusions, even demonstrably false ones, are "cognitively impenetrable" in that you can't stop seeing them when you know they're fake www.linkedin.com/posts/jonipp...
When we remark at AI's apparent intelligence, it's critical to remember that the Turing Test assesses the strength of an *illusion*. Some illusions, even…
Apparently there's minimal danger in this case, but can we please make techies legally responsible for blowback—or at least require them to document some thinking into the unintended consequences of their innovations? gizmodo.com/nasa-smashed... 🐘♻️
Rocks knocked off Dimorphos could come back to bite us in a few years—in the form of harmless meteors.
To keep student perspectives balanced, I've created a 7-minute video introducing the IMPACT RISK acronym for remembering ten AI downsides, from exploitation to automation to climate. Watch on YouTube www.youtube.com/watch?v=kUIb...bit.ly/impactrisk 🐘♻️
YouTube video by UMaineDigCuration
Iran is now using AI to polarize US elections. OpenAI found 5 websites and a dozen X/Instagram accounts that used ChatGPT to fabricate fake long-form stories and social media comments from progressive and conservative angles www.axios.com/2024/08/16/o... 🐘♻️
This is the first operation OpenAI has found and removed that focuses on the U.S. elections.
One of the under-recognized dangers of AI in the IMPACT RISK framework is M for Monopoly. The handful of companies dominating AI are now trying to evade antitrust scrutiny by siphoning out the guts of rival startups rather than buying them outright www.nytimes.com/2024/08/08/t... 🐘♻️
With the constant stream of news about bigger and better AI models, it can be hard to keep track of their dangers. So I made a free infographic to illustrate 10 downsides based on the IMPACT RISK mnemonic. Download a portrait-oriented PDF/PNG or slide for a presentation bit.ly/impactrisk 🐘♻️
Yeah, should be good (I'm not on it)!
While thinking about the AI metaphors panel today with @mahabali.bsky.social@annamillsoer.bsky.socialwww.linkedin.com/posts/jonipp... 🐘♻️
AI has made hiring more fraught than ever. This new arms race pits job-seekers, who've discovered that AI prefers Zoom candidates with bookshelf backgrounds, against employers, who are injecting prompts into job descriptions www.linkedin.com/posts/jonipp... 🐘♻️
AI has made hiring more fraught than ever. The new arms race pits job-seekers, who've discovered that AI prefers Zoom candidates with bookshelf backgrounds…