AI Reading for Saturday May 3
Get your free iris scan now. - Washington Post
Rant:
Thought experiment, what happens when anyone can alter video based on a prompt?
Tthere is video of everything in NYC now. So, someone gets assaulted, they gather video from neighbors (or cops but realistically only happens if someone got murdered), it goes to court.
Maybe you go to court, the defense attorney takes your video and puts Tom Cruise on it instead of the client, shows it and says, how do I know you didn't alter the video? So what has to happen is, cameras have to digitally sign video so you can show provenance and chain of evidence.
Anyone can ask AI to make a fake driver's license or passport. Reasonable ID verification requires you to look it up in government systems or a blockchain or something. But that becomes an online 'papiere bitte' situation.
It's crazy-making that the President goes and shows obviously Photoshopped stuff. Everybody should be able to open an image and see a trail of the digital signatures back to a specific camera tied to a specific media org. And nothing should really believed going forward without that really strong endorsement from reliable actors.
There's a huge surveillance and free speech problem. But there is also a huge problem with not being able to trust anything in a world where everything is mediated by Big Tech. I guess where I am now is, everything has to be digitally signed, including your posting, the infrastructure should be public, open-source, and standard, and there have to be bills of rights on under what circumstances video can be shared, people's IDs can be required, how they can be used by cops.
If you don't do universal standard secure infrastructure, you get flooded with bullshit by bad actors/morons like Trump, and you get corporate authoritarian islands, where you can't do anything unless your identity is vouched for by Big Tech.
You don’t want a world where you don’t exist and nothing is real unless Big Tech says so, Elon Musk can just put you in a ‘dead file’.
The problem of AI fakes and lack of trust is going to require a verified content and verified ID infrastructure, which requires a lot of privacy regulation and bills of rights to not become extremely authoritarian like China.
"NOOOO GROK IS WOKE!!" - Decrypt
Meta says capital spending could rise, blames Trump tariffs -The Register
Companies prepare for multiagent systems. - WSJ
Google teases NotebookLM mobile app in the Play Store ahead of I/O release - Ars Technica
Apple is working with Anthropic on an AI coding tool for Xcode - The Verge
Google rolls out AI Mode search to more users, adds features. - Google
How to Write a Resume Using AI - CNET
How Amazon red-teamed Alexa+ to keep your kids from ordering 50 pizzas - The Register
How I taught an AI to think like a painter. - FT
Airbnb deploys AI customer service bot, 50% of US users on board - Tech In Asia
re-upping this cartoon from yesterday
Robot goes berserk - Reddit
metaphor for our time
Follow the latest AI headlines via SkynetAndChill.com on Bluesky