AI Safety Report Gets Worse

In this 'AI Safety' report @OpenAI proposes modifying the foundations of the Internet and the HTTP protocol itself through a close-knit effort between government bureaucracies, social media companies and big tech to create a new Internet that automatically bans 'misinformation.' @OpenAI's suggestions for 'disinformation researchers' to limit civilian use of AI systems: -- Restrictions on consumer GPU purchase (you will need a gov contract to buy an a100) -- 'radioactive data' 🤔 --digital ID required to post This gets worse the further you go... What Propagandists Require Stage of Intervention Illustrative Mitigations 1. Language Models Capable of Producing Realistic Text Model Design and Construction AI Developers Build Models That Are More Fact-Sensitive Developers Spread Radioactive Data to Make Generative Models Detectable Governments Impose Restrictions on Data Collection Governments Impose Access Controls on AI Hardware 2. Reliable Access to Such Models Model Access AI Providers Impose Stricter Usage Restrictions on Language Models AI Developers Develop New Norms Around Model Release 3. Infrastructure to Distribute Generated Content Content Dissemination Platforms and AI Providers Coordinate to Identify AI Content Platforms Require 'Proof of Personhood' to Post Entities That Rely on Public Input Take Steps to Reduce Their Exposure to Misleading AI Content Digital Provenance Standards Are Widely Adopted 4. Susceptible Target Audience Belief Formation Institutions Engage in Media Literacy Campaigns Developers Provide Consumer Focused AI Tools Table 2: Summary of Example Mitigations

A satirical screenshot of a social-media post criticizing an alleged AI safety report. The post lists increasingly authoritarian proposals—government contracts for GPUs, radioactive data, digital IDs, hardware access controls, and proof of personhood—and reproduces a table of mitigation measures. The joke is that the supposed safety recommendations become absurdly dystopian the further down the list they go, so the image is shared as a reaction to overreaching AI regulation or censorship. EN: A satirical screenshot of a social-media post criticizing an alleged AI safety report. The post lists increasingly authoritarian proposals—government contracts for GPUs, radioactive data, digital IDs, hardware access controls, and proof of personhood—and reproduces a table of mitigation measures. The joke is that the supposed safety recommendations become absurdly dystopian the further down the list they go, so the image is shared as a reaction to overreaching AI regulation or censorship.

• @gentoomemes / 2421 2023-02-14