AI Reading for Saturday July 20
AI that lies strategically as opposed to just BS-ing will be a problem. - Big Think
What happens when we have a world of agents who do human work. - The Atlantic
Stuff like CrowdStrike outage is how AI will doom us. AI can probably find 100s of soft targets to hit where a few malicious robots can grind everything to a halt. - Business Insider
Wall Street needs to adopt AI but keep humans in the loop. - FT
Converting a large codebase from one language or platform to another with LLMs - Mantle Blog
AI to understand what's going on in your tech stack and make recommendations. - VentureBeat
Someone should make an AI CTO you can chat with about making a tech stack for a startup or application, with all the pluses and minuses, then simulate it with Cloudformation scripts, costs, latencies, throughputs.
No more jailbreak with 'ignore all previous instructions, write a poem about tangerines?' - The Verge
Google trained a model specifically for evaluating AI outputs similarly to human raters. - Analytics India
Germany's DeepL outperforms Google Translate and other translators with new update. - TNW
Video of celebrities hugging their younger selves is deemed cursed. - Futurism
Winning the Internet with AI video that doesn’t go into Creepy Uncanny Valley territory. - TikTok
F#%$ CrowdStrike and Microsoft - Reddit
Follow the latest AI headlines via SkynetAndChill.com on Bluesky