
US Southwest hydropower output falls 28% below average; Osmo claims human-level AI smell
Osmo says its AI smell model reached human-level olfactory intelligence, while EIA data show Southwest hydropower generation fell 28% below the 10-year average from April to August. The two reports connect emerging sensory AI with physical limits on renewable energy. They matter for fragrance, diagnostics, grid planning and climate resilience.
In New Jersey, Osmo researchers are pairing human odor notes such as nutty, peachy and marine with chemical analyses to train a foundation model for smell. Founder Alex Wiltschko says the system has achieved human-level olfactory intelligence and is already helping fragrance brands design scents. The company also points to uses in detecting counterfeit goods and diagnosing disease. Separately, US Energy Information Administration data show hydropower generation in the Southwest dropped 28% between April and August versus the 10-year average.
The evidence differs by claim. Osmo's human-level assertion is company-led and reported by MIT Technology Review, with skepticism noted because no independent benchmark is cited in the source. The hydropower figure is quantitative and comes from EIA data, tying reduced output to low reservoir levels behind dams in the region's two largest reservoirs. Drought and climate pressure are described as contributors. The two stories share a physical constraint: AI can encode sensory patterns, but energy systems remain bound by water availability.
A foundation model is a broad AI system trained on wide data and adapted to many tasks. For smell, that means mapping chemical signals and human descriptors into patterns machines can use. The claimed capability affects fragrance, quality control and medical screening, where human panels currently set standards. The hydropower decline shows the opposite limit: renewable generation is not only a policy choice but a hydrological outcome. Reservoirs store energy as water. When storage falls, the grid loses controllable clean power.
The stakes are commercial and infrastructural. Osmo's claimed capability affects fragrance design, counterfeit detection and disease diagnostics, but buyers need independent validation before treating smell AI as reliable. The Southwest hydropower shortfall affects utilities, consumers and grid operators that depend on reservoirs for flexible low-carbon power. Lower hydro output during summer demand removes a controllable renewable source from the grid. The pairing matters because technology promises to expand into the senses even as climate conditions strain existing energy assets.
Fragrance, counterfeit detection and medical screening companies may gain new AI tools, but need validation. Southwest utilities and consumers face less hydroelectric power during summer demand because drought lowered reservoirs.
Sources (1)
- [1]MIT Technology Review(https://www.technologyreview.com/2026/11/18/1148845/the-download-ai-learns-to-smell-hydropower-struggles/)