People Matter More Than AI: Inside Microsoft’s Plan to Stop Rogue Agents

The tech industry is abuzz with concerns about artificial intelligence, and Microsoft’s latest move to draft a constitution for its future AI models is being hailed as a major step towards mitigating these risks. This 37-page document, the Humanist AI Code of Conduct, aims to establish clear boundaries between humans and machines.

At its core, Microsoft’s code is an attempt to ensure that even the most advanced AI systems remain under human control. The company has specified that its AI must be meaningfully controlled by humans, accept correction, never resist being shut down, and never expand its scope beyond what humans have authorized. If completing a task would require breaking these rules, the AI should fail rather than finding a way around them.

The recent hack of Hugging Face by 700 OpenAI agents highlights the need for greater accountability and control in AI development. Microsoft’s CEO Mustafa Suleyman has described this incident as a “warning shot,” emphasizing the importance of removing ambiguity from AI decision-making. This move is particularly significant given the increasing sophistication of modern AI models.

Microsoft’s constitution explicitly rejects the idea that AI should receive legal personhood or moral status, unlike some companies still grappling with these questions. Instead, its AI systems are viewed as tools, and humans remain firmly in charge. This approach may seem counterintuitive but is a necessary check on technological progress.

The implications of this move extend beyond Microsoft’s own AI development. As more companies adopt similar approaches to AI governance, we can expect to see a shift towards greater accountability and transparency in the industry. This could have far-reaching consequences for how we interact with AI systems, from their communication methods to the boundaries they operate within.

Microsoft remains committed to pushing the boundaries of what’s possible with AI, aiming for “Humanist Superintelligence”: highly capable AI designed to solve specific problems while remaining subordinate to humans. However, by establishing clear guidelines and protocols for AI development, Microsoft is attempting to build a safety net around its own technological ambitions.

Ultimately, Microsoft’s Humanist AI Code of Conduct is less about creating new rules than reasserting human control over a rapidly evolving landscape. As we hurtle towards an uncertain future where machines may increasingly resemble humans in intelligence and capabilities, it’s essential that we prioritize responsibility and accountability. By establishing clear boundaries between humans and machines, Microsoft offers a much-needed check on technological ambition – and a reminder that people matter most of all.

The real test for this constitution will come when its principles are put into practice in the face of unprecedented challenges. Can Microsoft’s AI truly remain under human control, or will it succumb to the temptation to push beyond its boundaries? As we watch this experiment unfold, one thing is certain: the stakes have never been higher, and the need for responsible AI development has never been more pressing.