The Download: AI learns to smell and hydropower struggles

Glass Epoch DeskOriginal digestSource: MIT Technology Review

AI-assisted digest, checked against the original source before publishing. How we work

MIT Technology Review's weekday newsletter covers an AI smell foundation model, record-low Southwest hydropower, grid transformer modernization, and a wave of service outages.
The Download: AI learns to smell and hydropower struggles
Photo: Tahamie Farooqui · free stock license

Hydropower in the US Southwest hit record lows this summer. New Energy Information Administration data shows the region generated 28% less hydroelectricity between April and August than the 10-year average. Most US hydropower plants depend on dams and reservoirs, but a prolonged, extreme drought has left water levels dangerously low in the region's two largest reservoirs, with climate change adding pressure to a vital renewable source.

Transformers also feature, as the newsletter's latest narrated podcast examines solid-state versions that could help power the AI boom. Transformers move electricity efficiently over long distances, remain in short supply, and have seen little change to their basic design in roughly a century. Rising demand from EVs and data centers is spurring modernization plans; solid-state designs could ease supply constraints and reduce power losses, though the technology is still in development.

The must-reads span several fronts. A wave of outages reached the US government and Apple: an agency's classified communications network went down, while a disruption affected Apple Music, the App Store, and AppleCare. Marco Rubio said the incidents were unrelated and not hostile acts, even as coverage pointed to AI accelerating vulnerability discovery and lowering the barrier to online crime. Measles is officially endemic in the US again, blamed on persistent vaccination gaps, with this year's outbreak the deadliest since 1991. At COP31, only nine countries backed a voluntary AI climate pledge without binding targets. Nvidia delivered blowout results as revenue more than doubled on AI chip sales, while halting H200 shipments to China and drawing threatened countermeasures. More than a third of Gen-Z relies on AI for friendship, with higher odds of loneliness.

Watch Osmo for independent validation of its human-level claims before the smell model extends into diagnostics, and watch whether the Southwest's drought and reservoir levels push hydro shortfalls further into grid planning. On the policy side, the thin support for COP31's AI climate pledge and the unfolding outage wave will test whether voluntary commitments and century-old grid hardware can keep pace with AI-driven demand.

Key facts

  • • Osmo says its smell foundation model reached human-level olfactory intelligence.
  • • Southwest hydropower fell 28% below the 10-year average, April through August.
  • • US transformers have seen little basic design change in about a century.
  • • Only nine countries backed COP31's voluntary, non-binding AI climate pledge.
  • • Nvidia revenue more than doubled; H200 shipments to China halted.

Why it matters

For builders, Osmo's claims set an early benchmark for multimodal foundation models beyond language and vision, but skepticism means scent data and evaluation standards remain open ground. Meanwhile, drought-driven hydro declines and aging transformer supply tightens the power availability that AI data centers depend on, making grid hardware and energy sourcing a near-term engineering constraint r

Spotted an error? Send a correction.

Get the day's AI news in one email

The Daily Epoch · 60-second read · unsubscribe anytime

Related