DFO Pacific salmon AI — Chumputer, computer-vision migration counter, Factoid Finder — federal digital modernization counterweight to spending cuts
NA-Canada (Pacific — British Columbia)
Content
The Fisheries and Oceans Canada (DFO) 2026–27 Departmental Plan (June 2026) is explicit: “Explore options for policy and licensing changes that will modernize west coast commercial fisheries… Invest in digital modernization, including online licensing forms and pilot projects using artificial intelligence to digitize fishing data and reduce administrative burden.”
Three active DFO AI pilot projects in the Pacific Salmon Strategy Initiative (PSSI):
1. Chumputer vision AI — Pacific salmon scale-reading AI
Deep-learning CNNs that read salmon scale growth rings (circuli) to determine chum salmon age. Origin: manual age-reading by trained experts, 80,000+ scales/year, demand exceeds capacity. The Chumputer is the first step toward an all-salmon-species scale-reading AI. Next: Chinook, sockeye, coho.
Substantive observation. The Chumputer replaces a highly-trained-expert activity (scale-reading) with an AI. The 80,000+ scale figure is the operational scale of the manual process; the Chumputer is positioned to handle the same scale with greater throughput.
2. Computer-vision salmon migration counter
Automated fish counting at fish fences/weirs (Sproat and Stamp River fish ladders, Vancouver Island). Currently manual: trained staff count and identify sockeye/coho/Chinook from underwater camera footage, frame by frame. AI counters automate this in real time.
Substantive observation. The manual process is video-based expert counting — the AI replaces expert visual identification. The operational scale (frames per second, species accuracy) is the substantive metric; not publicly detailed.
3. Factoid Finder — AI tool for watershed planning
AI tool that “quickly reads, understands, and collates information from large literature collections” for the Integrated Planning for Salmon Ecosystems (IPSE) watershed planning process. Reads reports, scientific studies, policy documents. Used for the Nicola Watershed Integrated Salmon Ecosystem Strategy. Supports First Nations + Provincial/Territorial co-development.
Substantive observation. Factoid Finder is an NLP / LLM-style AI for environmental-policy literature review — distinct from CV salmon AI. The First Nations + Provincial/Territorial co-development framing is substantive.
DFO’s digital modernization counterweight
DFO’s 2026-27 spending reduction commitment: $54.47M (2026-27), $101.91M (2027-28), $193.82M (2028-29). Approximately 551 FTEs decrease by 2028-29. The AI pilots are part of the digital modernization counterweight to these cuts.
Substantive observation. The DFO pilots are cost-saving AI deployments in a federal department facing significant budget reductions. The pilots are positioned to preserve operational capacity as DFO loses FTEs. This is a substantive Canadian federal pattern — AI is being deployed in part to substitute for reduced human capacity rather than to augment it.
Cross-references — DFO Pacific salmon AI + Indigenous-led AI
The Salmon Vision deployment (BC Indigenous-led wild salmon AI) and PolArctic Sanikiluaq (Inuit-led mariculture AI) are operational Indigenous-led AI deployments. The DFO Chumputer / migration counter / Factoid Finder pilots are federal AI deployments. Together these form the Canadian Pacific + Northern salmon + mariculture AI ecosystem. The federal pilots and the Indigenous-led deployments are complementary, not redundant.
What this unit is doing in the taxonomy
This is the DFO Pacific salmon AI substrate unit. Anchors:
- The constraint/critical scout’s federal substrate analysis.
- The Canadian aquaculture AI unit (this scan).
- The PolArctic/Salmon Vision Indigenous-led AI observations.
- The existing Pacific/Atlantic fisheries AI vendor unit.
Why it matters for talks
- DFO Pacific salmon AI is the first documented federal regulator AI deployment with three named active pilots. Substantive corpus cell.
- The Chumputer is positioned to read 80,000+ scales/year — replacing the manual expert process with AI.
- Factoid Finder’s NLP/LLM-style AI for environmental-policy literature review is substantively distinct from CV salmon AI. The First Nations + Provincial/Territorial co-development framing is substantive.
- DFO’s digital modernization counterweight to ~$54M spending reduction is the substantive federal AI-as-substitute-for-FTE-reduction pattern. Worth carrying.
- The DFO + Indigenous-led AI (Salmon Vision + PolArctic) ecosystem is the substantive Canadian pattern — federal + Indigenous-led complementary, not redundant.
Critical context
- DFO’s three Pacific salmon AI pilots are pilot-stage. Operational scale (number of scales processed, number of fish counted, number of watershed plans supported) not publicly tracked. G-323.
- The Chumputer accuracy vs. trained-expert accuracy is not publicly documented.
- Factoid Finder’s literature review accuracy is not publicly documented.
- The 2026 DFO transition plan for 79 BC salmon farms remains unresolved as of July 2026 — operational future of Mowi Canada West Feed Centre + OnDeck Fisheries AI + Catalera-dependent salmon-farm AI deployments uncertain.
- The 551-FTE decrease by 2028-29 is the binding context for any AI-as-substitute-for-FTE-reduction framing.
- DFO’s $54.47M (2026-27) / $101.91M (2027-28) / $193.82M (2028-29) spending reduction is the binding context. AI pilots are positioned as the digital modernization counterweight.