Tag
Thinking Machines outlines its mission to build AI that extends human will and judgment by training strong models, building customizable tools, and developing interfaces that enable continuous human influence. The company argues for decentralized, human-shaped AI over frozen, centrally trained models.
The article explores how increasingly powerful AI models eliminate those whose skills can be encoded into prompts, emphasizing that the truly irreplaceable value lies in tacit knowledge, physical-world operations, and interpersonal trust. Through the example of a friend transitioning from a consultant to a hardware integrator, the author illustrates how proactively yielding to AI-replaceable tasks while deepening expertise in areas beyond AI's reach is key to surviving and thriving in the technological wave.
The article argues that agentic AI tools shift the bottleneck from coding ability to domain expertise, making those who can verify correctness in both code and domain the most valuable.
This paper introduces GrowLoop, a self-evolving evaluation system for assessing human-likeness in open-ended conversations. It uses minimal human seed annotations to iteratively refine evaluation rubrics, addressing challenges of tacit knowledge, varying human agreement, and evolving model capabilities.
An excerpt from a book-in-progress explores how Xerox repair technicians in the 1980s relied on social knowledge-sharing and storytelling ('war stories') to maintain complex photocopiers, based on anthropologist Julian Orr's ethnographic research published in 'Talking About Machines' (1996).