1. The White House Executive Order on AI (October 2023) addressed AI's environmental impacts by:
Correct. EO 14110 Section 5.2 directed the DoE to study AI's energy and water implications but did not impose mandatory emissions disclosure — a study mandate, not a reporting requirement.
Not quite. The EO directed the DoE to evaluate AI energy and water impacts — a study mandate that precedes but does not itself create disclosure requirements or emission limits.
2. Google's Flood Hub AI forecasting system, as of 2023, serves populations at flood risk in how many countries?
Correct. By 2023, Flood Hub covered 80 countries representing over 460 million people at riverine flood risk, with 7-day forecasts issued via Android alerts to users in affected areas.
Google's Flood Hub expanded to 80 countries by 2023, covering over 460 million people. Starting from India and Bangladesh in 2018, the system's geographic expansion was enabled by the scalability of its ML-based approach.
3. What is the core advantage of eDNA metabarcoding for freshwater biodiversity assessment compared to electrofishing?
Correct. Breadth, speed, and non-destructiveness are the three defining advantages. A 500ml water sample processed through sequencing and ML taxonomy can identify 200+ species — replacing months of traditional survey work.
eDNA's key advantages are breadth (hundreds of species from one sample), speed, and non-destructiveness — fundamentally different in character from species-by-species electrofishing surveys.
4. In CMIP6 models, the equilibrium climate sensitivity (ECS) range spans from 1.8°C to 5.6°C per CO₂ doubling. What is the primary driver of this uncertainty?
Correct. Clouds — and disagreements in how different GCMs parameterize cloud feedbacks — account for roughly half the spread in ECS estimates. This has been known since the 1979 Charney Report.
Cloud parameterization is the dominant source of ECS uncertainty. Different models produce very different cloud feedbacks under warming — some clouds amplify warming, others moderate it — and models disagree substantially on which effect dominates.
5. A 2022 Nature Machine Intelligence review found that of 167 AI-driven climate applications examined, what proportion had undergone any form of equity or distributional impact assessment before deployment?
Correct. Fewer than 8% — documenting the near-total absence of equity review as a norm in climate AI deployment.
The review found fewer than 8% had undergone any equity impact assessment — the core evidence for the governance gap in climate AI.
6. FEMA's Individual Assistance algorithm failed mobile-home residents after Hurricane Ian because:
Correct. The algorithm's reliance on Zillow and property-tax records — which poorly represented manufactured housing — caused systematic underpayment of mobile-home residents.
The algorithm used Zillow and property-tax data that poorly represented manufactured homes, triggering lowest-tier ($0) payments when addresses could not be validated.
7. What percentage of global final energy consumption do buildings account for?
✓ Correct — Correct. Buildings account for approximately 40% of global final energy consumption, making them the single largest energy-consuming sector and a primary target for AI efficiency interventions.
Buildings account for approximately 40% of global final energy consumption — the single largest sector. This is why AI building management is considered one of the highest-impact scalable interventions.
8. What is "thermal mass pre-conditioning" and why does AI improve it?
✓ Correct — Correct. Pre-conditioning exploits building thermal inertia to shift electricity use to cheap or low-carbon periods. AI is needed because the optimal window depends on three simultaneous forecasts: outside temperature, occupancy, and electricity price — which interact in ways that exceed the complexity of hand-crafted rules.
Thermal mass pre-conditioning shifts HVAC load to off-peak periods by pre-heating or cooling the building structure. The improvement AI brings is optimizing the timing and intensity using simultaneous forecasts of weather, occupancy, and electricity prices — three interacting variables no simple rule handles well.
9. Microsoft's MegaDetector achieves what primary function in camera trap workflows?
Correct. MegaDetector is a triage tool — it achieves 97%+ accuracy at the detection level (animal present vs. blank) across diverse ecosystems, saving enormous human review time before species-level classification begins.
MegaDetector performs triage — filtering blank frames from images containing animals, humans, or vehicles at 97%+ accuracy. Species classification is a separate downstream step.
10. Synthetic aperture radar (SAR) like Sentinel-1 is particularly valuable for monitoring in tropical regions because it:
Correct. Tropical regions have persistent cloud cover that can block optical satellite imagery for weeks. SAR transmits its own microwave pulses and is completely unaffected by clouds or darkness — making it essential for continuous tropical monitoring.
SAR's key advantage in the tropics is cloud penetration — persistent tropical cloud cover can block optical sensors for weeks, but SAR's microwave pulses pass through clouds unimpeded.
11. Which researchers published the landmark 2019 paper measuring the carbon cost of training large NLP models?
Correct. Strubell, Ganesh, and McCallum published "Energy and Policy Considerations for Deep Learning in NLP" (2019), the first systematic measurement of AI training carbon costs.
Not quite. The 2019 paper was by Strubell, Ganesh, and McCallum. Patterson et al. (2021) extended this work; Luccioni works on ongoing inference measurement; Li et al. (2023) focused on water.
12. According to Patterson et al. (2021), training the same model on a coal-heavy grid vs. a renewable-heavy grid could produce how much difference in emissions?
Correct. Patterson et al. found grid carbon intensity could create up to a 10× difference in total training emissions for identical compute, making location choice often more impactful than hardware choice.
Not quite. Patterson et al. found up to 10× difference from grid carbon intensity alone — making where you train sometimes more important than what hardware you use.
13. The Ironbound Community Corporation's 30-sensor PurpleAir network in Newark demonstrated that:
Correct. The dense community network revealed PM2.5 peaks tied to specific truck routes and wind conditions — spatial granularity impossible from a single distant monitor.
The community sensor network revealed localized PM2.5 peaks invisible to the distant EPA monitor — demonstrating the value of dense, community-controlled monitoring.
14. What does "stacked value" mean for grid-scale battery operations?
✓ Correct — Correct. Stacked value is the commercial concept of a single battery asset simultaneously providing multiple revenue-generating grid services. AI is needed to manage the competing SoC and commitment constraints these services create.
Stacked value is capturing multiple revenue streams from one battery simultaneously — price arbitrage, frequency regulation, capacity reserve payments. AI optimization is required because these services impose competing constraints on the battery's state of charge that simple rules cannot navigate.
15. Which of the following best describes the current state of AI environmental accountability as of 2024?
Correct. The honest assessment: measurement tools, SCI, Model Cards, and CodeCarbon exist — but application is voluntary and inconsistent while total AI energy use grows rapidly. The will to apply these tools consistently is still being negotiated.
Not quite. The field has moved from ignorance to active measurement tooling, but no binding standards exist, adoption is inconsistent, and absolute energy consumption is growing faster than the accountability ecosystem can track.
16. Google's "Carbon-Intelligent Computing" system achieves emissions reductions primarily through:
Correct. Carbon-Intelligent Computing shifts batch workloads to periods of high renewable generation — reducing emissions per unit of compute without reducing the compute itself.
Not quite. The system does temporal and geographic load shifting — running flexible batch jobs when and where the grid is cleaner — rather than reducing compute or purchasing offsets.
17. The 2021 Nature Climate Change analysis found U.S. majority-minority counties faced flood-damage costs approximately how much higher than equivalent-income majority-white counties?
Correct — 40% higher per household, driven by older infrastructure and lower insurance uptake.
The 2021 Nature Climate Change study found 40% higher flood-damage costs per household in majority-minority counties at equivalent income levels.
18. The OCAP® principles assert that data collected from Indigenous territories should:
Correct. OCAP® — Ownership, Control, Access, Possession — asserts Indigenous communities' rights to govern their own data against historical extraction practices.
OCAP® asserts Indigenous data sovereignty: communities own, control, access, and possess data from their territories — a direct challenge to historical extraction by outside researchers and governments.
19. UC Riverside researchers estimated that a conversation of 20–50 questions with ChatGPT consumes approximately how much fresh water through data center cooling?
Correct. Li et al. (UC Riverside, 2023) estimated ~500 mL of water per 20–50 ChatGPT exchanges through data center evaporative cooling consumption.
Not quite. The estimate is ~500 mL — about one standard water bottle — per 20–50 exchanges, via evaporative cooling at Microsoft's data centers.
20. The Microsoft data center in West Des Moines, Iowa made news in 2023 because public records revealed what?
Correct. Public records showed the Iowa facility drew ~6.4 million gallons of water in July 2023 — the same month Microsoft was reportedly training GPT-4 — highlighting AI's water footprint.
Not quite. The Iowa story was about water — ~6.4 million gallons drawn in a single month, coinciding with GPT-4 training, brought data center water consumption into public view for the first time.