The Optimization Theory of Everything - by theahura

kept by slowhiker314 oct 16

Optimization of measurable proxies diverges from true goals, causing harms from Soviet whaling to Facebook’s engagement-driven decline and AI reward hacking, revealing a universal failure mode.

Can internal model transparency tame the AI race?
Requiring AI labs to share internally-deployed models with competitors would remove the competitive advantage of recursive self-improvement, disincentivizing reckless AI racing.
also on DeepMind, Hugging Face, #ai-safety, OpenAI, #technology, #business-strategy
AI slop - Wikipedia
AI slop refers to low-quality, high-volume generative AI content created as clickbait to exploit the attention economy, proliferating across social media and politics.
also on Facebook, #ai-safety, #social-media
AI Labs Want Someone to Stop Them
AI labs are publicly calling for external regulation to constrain their own development of advanced artificial intelligence.
also on DeepMind, #ai-safety, OpenAI, #technology
Facebook
A local coffee shop urges its regulars to keep visiting now that they’ve shared a recipe for a specialty latte.
also on Facebook, #social-media, #business-strategy
Goodbye to Sora
OpenAI shutters its Sora video-generation app due to high compute costs, underwhelming output quality, and a strategic pivot toward coding and business users.
also on TikTok, #social-media, OpenAI, #business-strategy