← Observatory › Programmes › PROG-AI

Artificial Intelligence

The faultline

Whether artificial intelligence capabilities demonstrated under bounded conditions continue to hold beyond those conditions.

2026-10-01FR-AI-0006Normal Record Review admitted Keeman (2026), arXiv:2603.22295v1, as bounded evidence that affect-reception representations recur across six Llama/Gemma base and instruct variants while emotion categorisation improves with scale. Primary provenance recorded; no legacy instance altered.Fragmenting
2026-09-28FR-AI-0009IN-009 Project Kitchen / Game2Policy preprint added as qualified, task-specific game-to-robot transfer evidence. WIRED coverage and the Worldmodeldata–Mila planned study were not treated as demonstrated transfer results.Escalating
2026-09-26FR-AI-0002Normal Record Review of Frontline Scout flag 2026-09-26-01 admitted OSWorld-Pro arXiv:2609.24890v1 as bounded process-level contesting evidence at the low-supervision agentic boundary. AS-002 remains current; Pressure State ESCALATING, Verification Stage VS-02, mechanisms and open questions unchanged. The benchmark does not establish economic productivity or real-world supervision cost.Escalating
View Artificial Intelligence trajectoriesAI records are brought forward within the full archive.

No current programme-level diagnosis.