“AI #131 Part 2: Various Misaligned Things” by Zvi

It doesn’t look good, on many fronts, especially taking a stake in Intel. We continue. Table of Contents America Extorts 10% of Intel. Nice company you got there. Who's next? The Quest For No Regulations Whatsoever. a16z is at it again, Brockman joins. The Quest for Sane Regulations. Dean Ball surveys the state legislative landscape. Chip City. Nvidia beats earnings, Huawei plans to triple chip production. Once Again The Counterargument On Chip City. Sriram Krishnan makes a case. Power Up, Power Down. I for one do not think windmills are destroying America. People Really Do Not Like AI. Some dislike it more than others. A lot more. Did Google Break Their Safety Pledges With Gemini Pro 2.5? I think they did. Safety Third at xAI. Grok 4 finally has a model card. Better late than never. [...] --- First published: September 12th, 2025 Source: https://www.lesswrong.com/posts/CWYomHppHsxNe9xRC/ai-131-part-2-various-misaligned-things-1 --- Narrated by TYPE III AUDIO. ---Images from the article:<img src="https://substackcdn.com/image/fetch/$s_!PMKW!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F4bf46f7a-5485-4e8c-8d4b-dbbd47ad761b_825x709.png" alt="Bar graph showing QA performance comparison across six AI models. The graph titled "Simple QA no browse" compares performance metrics (Refusal, Incorrect, Correct) for Opus 4, Sonnet 4, OpenAI 03, OpenAI c4-mini, GPT-4b, and GPT-4.1." style="max-width: 100%;" /><img src="https://substackcdn.com/image/fetch/$s_!0Era!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F5b73e787-a98f-4424-b3ee-b2b354f35f8c_1098x759.png" alt="Bar graph titled "Overrefusal Scores" comparing AI models' inappropriate refusal rates. The graph shows scores for six different AI models (Claude Opus 4, Claude Sonnet 4, GPT-4.1, GPT-4.0, c3, and c4-mini), with c3 having notably higher overrefusal scores compared to the other models. The data includes 95% confidence intervals represented by error bars." style="max-width: 100%;" />Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.