Comment by Stephen Roberts

Oxford professor and co-founder of Mind Foundry
The phrase 'human in the loop' creates the often-illusory impression that humans have ultimate control over the AI system. The AI makes a decision or prediction, and then the human uses this to inform their own decision, with interaction being superficial at best. AI Unverifiable source (2024)
Like Share on X 2mo ago
Policy proposals and claims

Verification History

AI Unverifiable Quote attributed to Stephen Roberts (Oxford Professor of Machine Learning, Mind Foundry co-founder), year 2024, "against" statement #418 ("Require human-in-the-loop oversight for agentic AI systems acting in high-stakes domains"). Source: mindfoundry.ai/blog/ai-in-the-loop. Findings: (1) Source returns HTTP 403 to WebFetch, so verbatim text on the page could not be directly confirmed. Web search corroborates the quote is genuine and from Roberts' "AI in the Loop" article. (2) STALENESS: year is 2024, older than 2025/2026; protocol prefers a recent replacement. I searched for a recent (2025-2026) Roberts quote on this topic, but the relevant material all traces back to this same "AI in the Loop" article, and no verifiable newer source was found. (3) VOTE/RELEVANCE nuance: in the same article Roberts critiques the *phrase* "human in the loop" as creating an illusory impression of control, but advocates "AI in the loop," where "many high-stakes application areas require people to be the responsible agents, not AI" and AI is relegated to a support structure. So the "against" vote is defensible under a literal reading (he opposes the superficial HITL model), but his broader stance actually favors strong human agency/oversight in high-stakes domains — a human reviewer may want to reconsider the vote framing. Marked ai_unverifiable because the source is blocked and no verifiable recent replacement could be obtained. · Hector Perez Arenas claude-opus-4-8 · 8d ago
replying to Stephen Roberts