Tsamados et al. 2024 state that the teaming paradigm 'rests on the assumption that the pace of innovation in AI research remains centred on foundation models,' and that with more explainable, predictable architectures 'supervisory human control may re-emerge as the primary approach to controlling AI systems' — the shift from supervision to teaming is a bet on a technical trajectory, not a settled endpoint.
E48 contains both quotes verbatim and explicitly conditions HMT's promise on the foundation-model innovation trajectory, with supervisory control possibly re-emerging under more explainable architectures — directly implying the 'bet on a technical trajectory, not a settled endpoint' reading.
Written by Kimi K3 via Ollama Cloud · checked by GLM-5.3 via Ollama Cloud · 30 Oct, 04:57
Source chain
AEvery quote below was checked, without a model, to appear verbatim in its source.
- 01
“supervisory human control may re-emerge as the primary approach to controlling AI systems”
Full-text passage · no page number · evidence E48
The hypothesis that HMT can offer a productive path forward for the human control of AI systems rests on the assumption that the pace of innovation in AI research remains centred on foundation models instead of data and compute-efficient architectures that produce more explainable and predictable AI systems (e.g., neuro-symbolic AI). In the event of a departure from the non-deterministic and inscrutable models, we described in this analysis, supervisory human control may re-emerge as the primary approach to controlling AI systems. Nevertheless, as our analysis contends, the benefit of exploring HMT remains useful and transferrable to future human–AI configurations as it explores AI systems' cooperative and collaborative capabilities, the expectations of human operators, and the characteristics of an effective multi-agent system. Research investment into HMT and collaborative agency can be beneficial in the long term, irrespective of a change in AI capabilities, as it focuses on approaches that improve the quality of human–AI interactions and co-action.
Passage read from www.ebi.ac.uk, which may be a preprint rather than the published version.