RT Liron Shapira (@liron)Doom Debate with @VitalikButerin has dropped:Will “d/acc” Protect Humanity from Superintelligent AI?Vitalik is both a hardcore techno-optimist and builder, and someone willing to seriously consider AI regulation and coordination mechanisms. He coined the term "d/acc'" – defensive, decentralized acceleration – as a middle path between uncritical AI acceleration and total pause scenarios.We debate the tractability of AI alignment, whether approaches like d/acc would actually work when ...

Core Insights - The discussion at the 2025 World Artificial Intelligence Conference (WAIC) highlighted the urgent need for global AI governance and the challenges associated with defining usage boundaries in AI technology [1][4]. AI Development in China - Eric Schmidt emphasized that the primary challenge in global AI governance is not model access but the consensus on usage boundaries, indicating a need for international coordination to establish enforceable safety mechanisms [4][6]. US-China AI Cooperation - Schmidt argued that US-China cooperation on AI must be based on shared objectives, particularly regarding high-risk issues like AI-controlled weapons and autonomous learning, advocating for dialogue to ensure human control over AI developments [5][7]. Technological Openness vs. Security - Schmidt noted that while open-source approaches in Chinese AI models foster innovation, they also increase vulnerability to misuse, suggesting that global collaboration is essential to refine protective mechanisms without abandoning openness [6][8]. Competition and Cooperation - Reflecting on the competitive dynamics among major tech companies, Schmidt stated that similar principles apply to nations, advocating for US-China cooperation on AI governance to maintain global stability and prevent conflict [7][8]. Ethical Foundations of AI Governance - The conversation underscored the importance of a shared moral framework in AI governance, echoing themes from Schmidt's co-authored book, warning that without it, humanity risks losing control over technological advancements [8].

The world if "AI alignment" people worked on spam control this whole time instead https://t.co/o35Ul0FzZM ...

Former Intel CEO launches a benchmark to measure AI alignment | TechCrunch https://t.co/cbe7RxK0A1 ...

My response to AI 2027:https://t.co/hqqXLM5UZlThe AI 2027 post is high quality, I encourage people to read it at https://t.co/84O5mfN5dhI argue a misaligned AI will not be able to win nearly as easily as the AI 2027 scenario assumes, because it greatly underrates our ability to protect ourselves, especially given the (pretty magical) technologies that the authors admit will be available in 2029 in their scenario. ...