{ "id": "AIBD-0164", "title": "DeepMind reinforcement-learning controller shapes TCV plasmas, including sustained twin “droplets”", "summary": "A deep RL controller commanding all 19 TCV coils produced and held elongated, negative-triangularity, snowflake and twin-droplet plasmas at EPFL's Swiss Plasma Center (Nature, 16 February 2022).", "date_observed": "2022-02-16", "date_basis": "unassessed", "date_precision": "unassessed", "benefit_status": null, "source_collection": null, "source_entry": null, "account": "Degrave et al. (Nature, published 16 February 2022; received 14 July 2021) describe a single deep neural network controller. It learns in simulation to command all 19 magnetic control coils of the Tokamak à Configuration Variable (TCV). TCV is a research tokamak at the Swiss Plasma Center, EPFL. Deployed on the real device, the controller produced and controlled elongated conventional shapes and advanced configurations such as negative triangularity and “snowflake” plasmas. It tracked location, current and shape. In one demonstration it established a snowflake with X-points 34 cm apart and narrowed the separation to 6.6 cm, combined RMSE 3.7 cm. It also sustained “droplets”, two separate plasmas maintained simultaneously in the vessel. MIT Technology Review (16 February 2022) reported that the AI controlled the real reactor without additional fine-tuning. It quoted independent expert Jonathan Citrin of the Dutch Institute for Fundamental Energy Research and Swiss Plasma Center director Ambrogio Fasoli. The experiment dates are not given.", "category": "benefit created", "unexpected": false, "purpose": "The controller was built to shape and maintain plasmas in TCV from high-level objectives; the configurations produced, including droplets, were its intended use.", "system": "DeepMind deep reinforcement learning magnetic controller for TCV (MLP policy commanding 19 coils)", "model_version": null, "developer": "DeepMind, with the Swiss Plasma Center, EPFL", "deployer": "Swiss Plasma Center, EPFL (TCV)", "location": "Lausanne, Switzerland", "domain": "Energy research (magnetic confinement fusion)", "segment": "Tokamak plasma control", "benefit_type": "Improved Task Accuracy", "taxonomy_note": "The result is a control capability on a research tokamak rather than an accuracy gain over a baseline.", "registry_ids": {}, "ratings": [ { "group": "organizational", "depth": 3, "reason": "The Swiss Plasma Center gained a controller the paper calls a notable advance for tokamak feedback control. It brought less design effort for new configurations and new droplet experiments." }, { "group": "societal", "depth": null, "reason": "No societal outcome beyond research progress is documented." } ], "ai_role": { "status": "assessed", "statement": "The reinforcement learning policy directly issued the coil voltage commands during the discharges, so the configurations are its output. The evidence is the peer-reviewed paper by the maker (DeepMind) and the operator (EPFL) plus an independent news report quoting an outside expert. Developer and deployer collaborated on the paper." }, "evidence": [ { "kind": "peer-reviewed study", "citation": "Degrave, J., Felici, F., Buchli, J. et al., 'Magnetic control of tokamak plasmas through deep reinforcement learning', Nature 602, 414–419 (2022), doi:10.1038/s41586-021-04301-9", "title": "Magnetic control of tokamak plasmas through deep reinforcement learning", "url": "https://www.nature.com/articles/s41586-021-04301-9", "date": "2022-02-16", "written_by": "makers", "quote": "We also demonstrate sustained ‘droplets’ on TCV, in which two separate plasmas are maintained simultaneously within the vessel." }, { "kind": "news report", "citation": "Will Douglas Heaven, 'DeepMind's AI can control superheated plasma inside a fusion reactor', MIT Technology Review, 16 February 2022", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "the researchers gave it control of the magnets in the Variable Configuration Tokamak (TCV), an experimental reactor in Lausanne." }, { "kind": "news report", "citation": "MIT Technology Review, 16 February 2022 (Jonathan Citrin, Dutch Institute for Fundamental Energy Research)", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "It's an incredibly powerful method. It's an important first step in a very exciting direction." }, { "kind": "preprint", "citation": "Tracey, B. D., Michi, A., Chervonyi, Y., et al. (2023). Towards practical reinforcement learning for tokamak magnetic control. arXiv:2307.11546, 21 July 2023 (v2 5 October 2023).", "title": "Towards practical reinforcement learning for tokamak magnetic control", "url": "https://arxiv.org/abs/2307.11546", "date": "2023-07-21", "written_by": "makers", "quote": null } ], "replication": { "produced_by": [ "A high-fidelity simulator (FGE) and equilibrium reconstruction (LIUQE) of the target tokamak available for training", "A flexible research tokamak with 19 independently actuated coils", "A real-time control system able to run a neural network policy at the required rate" ], "must_hold": [ "A simulator accurate enough for zero-shot transfer to the device", "Operator-specified targets within the device's physical and operational constraints", "Access to machine time on a research tokamak" ], "echoes": [ { "date": "2024-02-21", "setting": "DIII-D tokamak, San Diego; deep RL controller for tearing-instability avoidance (different objective)", "model_version": null, "outcome": "recurred", "source": "https://www.nature.com/articles/s41586-024-07024-9" } ] }, "paired_incident": null, "headline": "DeepMind RL controller shapes TCV plasmas and sustains droplets", "metric": "narrowed the separation to 6.6 cm with a combined RMSE of 3.7 cm", "comparator": null, "later": { "status": "persisted", "date": "2023-07-21", "note": "DeepMind and the Swiss Plasma Center reported upgraded RL controllers with up to 65% better shape accuracy in simulation and new validating experiments on TCV.", "quote": "We present simulation results that show up to 65\\% improvement in shape accuracy, achieve substantial reduction in the long-term bias of the plasma current, and additionally reduce the training time required to learn new tasks by a factor of 3 or more.", "url": "https://arxiv.org/abs/2307.11546" }, "value": null, "developer_country": "United Kingdom", "deployer_country": "Switzerland", "unrated_because": null, "brief_notes": null, "brief": null, "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "source": null, "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Improved Task Accuracy under Trustible's full taxonomy of 23 types; it was Improved Knowledge Discovery among the six first used.", "source": "https://trustible.ai/post/a-framework-for-measuring-the-benefits-of-ai/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Added what happened since from the cited sources.", "source": "https://arxiv.org/abs/2307.11546", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Country of the developer and the deployer added from the organization's own site or a registry.", "source": "https://en.wikipedia.org/wiki/Google_DeepMind; https://www.epfl.ch/research/domains/swiss-plasma-center/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "source": null, "editor": "AIBD" } ], "confidence": "reported", "shade": "medium", "claim_status": "Independently reported", "claim_basis": "Sources: one independent news report, the makers' peer-reviewed study and the makers' preprint. The cited evidence includes no independent study or report and no finding by a regulator or court.", "profile": { "case_id": "AIBD-0164", "catalog_version": "1.0.0", "source_fingerprint": "067f32980f20d353aff915cf6761f2d4e63e74512577a32df918c8220d5821b7", "collection_refs": [ { "collection_id": "aibd", "version": "snapshot-8bf7cfec9cb185aa" }, { "collection_id": "aibd", "version": "2026-10-25" }, { "collection_id": "viator", "version": "2026-10-25" } ], "source_ids": [ "AIBD-0164-source-0001", "AIBD-0164-source-0002", "AIBD-0164-source-0003", "AIBD-0164-source-0004" ], "families": { "strategic assessments": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] }, "evaluations": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] }, "monetary observations": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] }, "institutional decisions": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "record_ids": [] } }, "fields": { "actors[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "developer": "DeepMind, with the Swiss Plasma Center, EPFL", "deployer": "Swiss Plasma Center, EPFL (TCV)" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "developer", "deployer" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "actors[].ownership_control": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "ai_contribution": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ai_role": { "status": "assessed", "statement": "The reinforcement learning policy directly issued the coil voltage commands during the discharges, so the configurations are its output. The evidence is the peer-reviewed paper by the maker (DeepMind) and the operator (EPFL) plus an independent news report quoting an outside expert. Developer and deployer collaborated on the paper." } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ai_role" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "ai_relationship": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "category": "benefit created", "ai_role": { "status": "assessed", "statement": "The reinforcement learning policy directly issued the coil voltage commands during the discharges, so the configurations are its output. The evidence is the peer-reviewed paper by the maker (DeepMind) and the operator (EPFL) plus an independent news report quoting an outside expert. Developer and deployer collaborated on the paper." } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ai_role", "category" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].alternatives": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].assessor_date": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].depth": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "depth": 3, "reason": "The Swiss Plasma Center gained a controller the paper calls a notable advance for tokamak feedback control. It brought less design effort for new configurations and new droplet experiments." }, { "group": "societal", "reason": "No societal outcome beyond research progress is documented." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].display_confidence": { "state": "value_present", "value": { "kind": "structured", "data": { "confidence": "medium", "evidence_word": "reported", "claim_status": "Independently reported", "basis": "Sources: one independent news report, the makers' peer-reviewed study and the makers' preprint. The cited evidence includes no independent study or report and no finding by a regulator or court.", "interpretation": "Existing display rule preserved; the richer outcome-specific confidence question still needs review." } }, "provenance": { "kind": "curatorial", "basis": "Derived from the existing case display properties; not a new evidentiary judgment.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].rationale": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "depth": 3, "reason": "The Swiss Plasma Center gained a controller the paper calls a notable advance for tokamak feedback control. It brought less design effort for new configurations and new droplet experiments." }, { "group": "societal", "reason": "No societal outcome beyond research progress is documented." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "assessments[].rubric_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].status_basis": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "assessments[].target": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "depth": 3, "reason": "The Swiss Plasma Center gained a controller the paper calls a notable advance for tokamak feedback control. It brought less design effort for new configurations and new droplet experiments." }, { "group": "societal", "reason": "No societal outcome beyond research progress is documented." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "availability": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "brief": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "case_kind": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "case_links[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].evidence_links": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].evidence_strength": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].missingness": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "claims[].proposition": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "cross_border_flows[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "decisions[].action_outcome": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].grounds": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].institution_task": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "decisions[].success_criterion": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "institutional decisions" }, "deployment_stage": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "economics[].amount": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].costs[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].counterfactual": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].currency_price_basis": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].gross_net": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].measure_type": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].period_unit": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].realization": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].recipient_payer": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].return_payback": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "economics[].valuation_method": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "monetary observations" }, "evaluations[].ai_uplift": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].comparator": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].design": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].effect": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].exposure": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].id": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].measurement_method": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].metric": { "state": "not_assessed", "provenance": { "kind": "legacy_record", "basis": "Mapped from the existing case description without a new external source review.", "record_fields": [ "metric" ], "source_refs": [] }, "review_state": "not_assessed", "coverage": "partial", "family": "evaluations", "display": { "answer": "narrowed the separation to 6.6 cm with a combined RMSE of 3.7 cm", "editor_note": "" } }, "evaluations[].reproduction_materials": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].setting_sample": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "evaluations[].uncertainty_limitations": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "evaluations" }, "external_references[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "follow_up[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "later": { "status": "persisted", "date": "2023-07-21", "note": "DeepMind and the Swiss Plasma Center reported upgraded RL controllers with up to 65% better shape accuracy in simulation and new validating experiments on TCV.", "quote": "We present simulation results that show up to 65\\% improvement in shape accuracy, achieve substantial reduction in the long-term bias of the plasma current, and additionally reduce the training time required to learn new tasks by a factor of 3 or more.", "url": "https://arxiv.org/abs/2307.11546" } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "later" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "human_oversight": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "id": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "id": "AIBD-0164" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "id" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "infrastructure_dependencies[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].beneficiaries": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].beneficiary_count": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].benefit_type": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "benefit_type": "Improved Task Accuracy", "taxonomy_note": "The result is a control capability on a research tokamak rather than an accuracy gain over a baseline." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "benefit_type", "taxonomy_note" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].claim": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "account": "Degrave et al. (Nature, published 16 February 2022; received 14 July 2021) describe a single deep neural network controller. It learns in simulation to command all 19 magnetic control coils of the Tokamak à Configuration Variable (TCV). TCV is a research tokamak at the Swiss Plasma Center, EPFL. Deployed on the real device, the controller produced and controlled elongated conventional shapes and advanced configurations such as negative triangularity and “snowflake” plasmas. It tracked location, current and shape. In one demonstration it established a snowflake with X-points 34 cm apart and narrowed the separation to 6.6 cm, combined RMSE 3.7 cm. It also sustained “droplets”, two separate plasmas maintained simultaneously in the vessel. MIT Technology Review (16 February 2022) reported that the AI controlled the real reactor without additional fine-tuning. It quoted independent expert Jonathan Citrin of the Dutch Institute for Fundamental Energy Research and Swiss Plasma Center director Ambrogio Fasoli. The experiment dates are not given.", "benefit_type": "Improved Task Accuracy" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "account", "benefit_type" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].distribution": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].duration": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].id": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].stakeholder_group": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "ratings": [ { "group": "organizational", "depth": 3, "reason": "The Swiss Plasma Center gained a controller the paper calls a notable advance for tokamak feedback control. It brought less design effort for new configurations and new droplet experiments." }, { "group": "societal", "reason": "No societal outcome beyond research progress is documented." } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "ratings" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "outcomes[].status": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].tradeoffs": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "outcomes[].unexpected": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "unexpected": false, "purpose": "The controller was built to shape and maintain plasmas in TCV from high-level objectives; the configurations produced, including droplets, were its intended use." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "unexpected", "purpose" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "places[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "location": "Lausanne, Switzerland", "developer_country": "United Kingdom", "deployer_country": "Switzerland" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "location", "developer_country", "deployer_country" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "purpose": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "purpose": "The controller was built to shape and maintain plasmas in TCV from high-level objectives; the configurations produced, including droplets, were its intended use." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "purpose" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "record_dates": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "date_observed": "2022-02-16", "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Improved Task Accuracy under Trustible's full taxonomy of 23 types; it was Improved Knowledge Discovery among the six first used.", "source": "https://trustible.ai/post/a-framework-for-measuring-the-benefits-of-ai/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Added what happened since from the cited sources.", "source": "https://arxiv.org/abs/2307.11546", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Country of the developer and the deployer added from the organization's own site or a registry.", "source": "https://en.wikipedia.org/wiki/Google_DeepMind; https://www.epfl.ch/research/domains/swiss-plasma-center/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "editor": "AIBD" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "date_observed", "revisions" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "replication.attempts[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "A high-fidelity simulator (FGE) and equilibrium reconstruction (LIUQE) of the target tokamak available for training", "A flexible research tokamak with 19 independently actuated coils", "A real-time control system able to run a neural network policy at the required rate" ], "must_hold": [ "A simulator accurate enough for zero-shot transfer to the device", "Operator-specified targets within the device's physical and operational constraints", "Access to machine time on a research tokamak" ], "echoes": [ { "date": "2024-02-21", "setting": "DIII-D tokamak, San Diego; deep RL controller for tearing-instability avoidance (different objective)", "outcome": "recurred", "source": "https://www.nature.com/articles/s41586-024-07024-9" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "replication.conditions": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "replication": { "produced_by": [ "A high-fidelity simulator (FGE) and equilibrium reconstruction (LIUQE) of the target tokamak available for training", "A flexible research tokamak with 19 independently actuated coils", "A real-time control system able to run a neural network policy at the required rate" ], "must_hold": [ "A simulator accurate enough for zero-shot transfer to the device", "Operator-specified targets within the device's physical and operational constraints", "Access to machine time on a research tokamak" ], "echoes": [ { "date": "2024-02-21", "setting": "DIII-D tokamak, San Diego; deep RL controller for tearing-instability avoidance (different objective)", "outcome": "recurred", "source": "https://www.nature.com/articles/s41586-024-07024-9" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "revision_history[]": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "revisions": [ { "date": "2026-10-20", "change": "Record drafted from the cited sources.", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Benefit type set to Improved Task Accuracy under Trustible's full taxonomy of 23 types; it was Improved Knowledge Discovery among the six first used.", "source": "https://trustible.ai/post/a-framework-for-measuring-the-benefits-of-ai/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Added what happened since from the cited sources.", "source": "https://arxiv.org/abs/2307.11546", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Country of the developer and the deployer added from the organization's own site or a registry.", "source": "https://en.wikipedia.org/wiki/Google_DeepMind; https://www.epfl.ch/research/domains/swiss-plasma-center/", "editor": "AIBD" }, { "date": "2026-10-21", "change": "Prose pass: a rating reason, account, the system's part rewritten in shorter sentences with abbreviations expanded; facts, figures and quotations unchanged.", "editor": "AIBD" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "revisions" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sector": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "domain": "Energy research (magnetic confinement fusion)", "segment": "Tokamak plasma control" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "domain", "segment" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].access_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "sources[].citation_locator": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "evidence": [ { "kind": "peer-reviewed study", "citation": "Degrave, J., Felici, F., Buchli, J. et al., 'Magnetic control of tokamak plasmas through deep reinforcement learning', Nature 602, 414–419 (2022), doi:10.1038/s41586-021-04301-9", "title": "Magnetic control of tokamak plasmas through deep reinforcement learning", "url": "https://www.nature.com/articles/s41586-021-04301-9", "date": "2022-02-16", "written_by": "makers", "quote": "We also demonstrate sustained ‘droplets’ on TCV, in which two separate plasmas are maintained simultaneously within the vessel." }, { "kind": "news report", "citation": "Will Douglas Heaven, 'DeepMind's AI can control superheated plasma inside a fusion reactor', MIT Technology Review, 16 February 2022", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "the researchers gave it control of the magnets in the Variable Configuration Tokamak (TCV), an experimental reactor in Lausanne." }, { "kind": "news report", "citation": "MIT Technology Review, 16 February 2022 (Jonathan Citrin, Dutch Institute for Fundamental Energy Research)", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "It's an incredibly powerful method. It's an important first step in a very exciting direction." }, { "kind": "preprint", "citation": "Tracey, B. D., Michi, A., Chervonyi, Y., et al. (2023). Towards practical reinforcement learning for tokamak magnetic control. arXiv:2307.11546, 21 July 2023 (v2 5 October 2023).", "title": "Towards practical reinforcement learning for tokamak magnetic control", "url": "https://arxiv.org/abs/2307.11546", "date": "2023-07-21", "written_by": "makers" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "evidence" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].id": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "AIBD-0164-source-0001", "AIBD-0164-source-0002", "AIBD-0164-source-0003", "AIBD-0164-source-0004" ], "role": "Existing evidence items, in their original order" }, "provenance": { "kind": "curatorial", "basis": "Stable within the fingerprinted record version; no source content or source review has been added.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "complete" }, "sources[].language": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "sources[].relationship": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "evidence": [ { "kind": "peer-reviewed study", "citation": "Degrave, J., Felici, F., Buchli, J. et al., 'Magnetic control of tokamak plasmas through deep reinforcement learning', Nature 602, 414–419 (2022), doi:10.1038/s41586-021-04301-9", "title": "Magnetic control of tokamak plasmas through deep reinforcement learning", "url": "https://www.nature.com/articles/s41586-021-04301-9", "date": "2022-02-16", "written_by": "makers", "quote": "We also demonstrate sustained ‘droplets’ on TCV, in which two separate plasmas are maintained simultaneously within the vessel." }, { "kind": "news report", "citation": "Will Douglas Heaven, 'DeepMind's AI can control superheated plasma inside a fusion reactor', MIT Technology Review, 16 February 2022", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "the researchers gave it control of the magnets in the Variable Configuration Tokamak (TCV), an experimental reactor in Lausanne." }, { "kind": "news report", "citation": "MIT Technology Review, 16 February 2022 (Jonathan Citrin, Dutch Institute for Fundamental Energy Research)", "title": "DeepMind's AI can control superheated plasma inside a fusion reactor", "url": "https://www.technologyreview.com/2022/02/16/1045470/deepminds-ai-can-control-superheated-plasma-inside-a-fusion-reactor/", "date": "2022-02-16", "written_by": "independent", "quote": "It's an incredibly powerful method. It's an important first step in a very exciting direction." }, { "kind": "preprint", "citation": "Tracey, B. D., Michi, A., Chervonyi, Y., et al. (2023). Towards practical reinforcement learning for tokamak magnetic control. arXiv:2307.11546, 21 July 2023 (v2 5 October 2023).", "title": "Towards practical reinforcement learning for tokamak magnetic control", "url": "https://arxiv.org/abs/2307.11546", "date": "2023-07-21", "written_by": "makers" } ] } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "evidence" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "sources[].upstream_sources": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "strategic_assessments[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial", "family": "strategic assessments" }, "summary": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "summary": "A deep RL controller commanding all 19 TCV coils produced and held elongated, negative-triangularity, snowflake and twin-droplet plasmas at EPFL's Swiss Plasma Center (Nature, 16 February 2022).", "account": "Degrave et al. (Nature, published 16 February 2022; received 14 July 2021) describe a single deep neural network controller. It learns in simulation to command all 19 magnetic control coils of the Tokamak à Configuration Variable (TCV). TCV is a research tokamak at the Swiss Plasma Center, EPFL. Deployed on the real device, the controller produced and controlled elongated conventional shapes and advanced configurations such as negative triangularity and “snowflake” plasmas. It tracked location, current and shape. In one demonstration it established a snowflake with X-points 34 cm apart and narrowed the separation to 6.6 cm, combined RMSE 3.7 cm. It also sustained “droplets”, two separate plasmas maintained simultaneously in the vessel. MIT Technology Review (16 February 2022) reported that the AI controlled the real reactor without additional fine-tuning. It quoted independent expert Jonathan Citrin of the Dutch Institute for Fundamental Energy Research and Swiss Plasma Center director Ambrogio Fasoli. The experiment dates are not given." } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "summary", "account" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "systems[].autonomy_permissions": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].component_role": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].deployment_configuration": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].inputs_outputs": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].model_name": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].model_version": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "systems[].product_name": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "system": "DeepMind deep reinforcement learning magnetic controller for TCV (MLP policy commanding 19 coils)" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "system" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "systems[].provider": { "state": "value_present", "value": { "kind": "structured", "data": { "application_developer": "DeepMind, with the Swiss Plasma Center, EPFL" } }, "provenance": { "kind": "legacy_record", "basis": "Mapped from the existing case description without a new external source review.", "record_fields": [ "developer" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "taxonomy_crosswalks[]": { "state": "not_assessed", "review_state": "not_assessed", "coverage": "partial" }, "title": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "title": "DeepMind reinforcement-learning controller shapes TCV plasmas, including sustained twin “droplets”", "headline": "DeepMind RL controller shapes TCV plasmas and sustains droplets" } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "title", "headline" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" }, "version": { "state": "value_present", "value": { "kind": "identifiers", "identifiers": [ "067f32980f20d353aff915cf6761f2d4e63e74512577a32df918c8220d5821b7" ], "role": "SHA-256 of normalized original case data" }, "provenance": { "kind": "curatorial", "basis": "Deterministic content version assigned by the local migration.", "record_fields": [], "source_refs": [] }, "review_state": "record_mapped", "coverage": "complete" }, "workflow": { "state": "value_present", "value": { "kind": "record_fields", "fields": { "account": "Degrave et al. (Nature, published 16 February 2022; received 14 July 2021) describe a single deep neural network controller. It learns in simulation to command all 19 magnetic control coils of the Tokamak à Configuration Variable (TCV). TCV is a research tokamak at the Swiss Plasma Center, EPFL. Deployed on the real device, the controller produced and controlled elongated conventional shapes and advanced configurations such as negative triangularity and “snowflake” plasmas. It tracked location, current and shape. In one demonstration it established a snowflake with X-points 34 cm apart and narrowed the separation to 6.6 cm, combined RMSE 3.7 cm. It also sustained “droplets”, two separate plasmas maintained simultaneously in the vessel. MIT Technology Review (16 February 2022) reported that the AI controlled the real reactor without additional fine-tuning. It quoted independent expert Jonathan Citrin of the Dutch Institute for Fundamental Energy Research and Swiss Plasma Center director Ambrogio Fasoli. The experiment dates are not given.", "replication": { "produced_by": [ "A high-fidelity simulator (FGE) and equilibrium reconstruction (LIUQE) of the target tokamak available for training", "A flexible research tokamak with 19 independently actuated coils", "A real-time control system able to run a neural network policy at the required rate" ], "must_hold": [ "A simulator accurate enough for zero-shot transfer to the device", "Operator-specified targets within the device's physical and operational constraints", "Access to machine time on a research tokamak" ], "echoes": [ { "date": "2024-02-21", "setting": "DIII-D tokamak, San Diego; deep RL controller for tearing-instability avoidance (different objective)", "outcome": "recurred", "source": "https://www.nature.com/articles/s41586-024-07024-9" } ] } } }, "provenance": { "kind": "legacy_record", "basis": "Preserved from the existing AIBD record; external sources were not rechecked.", "record_fields": [ "account", "replication" ], "source_refs": [] }, "review_state": "record_mapped", "coverage": "partial" } }, "summary": { "total": 91, "addressed": 91, "known": 29, "reviewed": 0, "record_mapped": 29, "not_assessed": 62, "complete": false, "by_state": { "value_present": 29, "not_reported": 0, "unresolved": 0, "not_assessed": 62, "not_applicable": 0, "withheld": 0, "source_unavailable": 0 } } } }