Today, look at how researchers, engineers and communities approach these problems.

Then turn one idea into a project you could begin.

Which decision from yesterday would you most like to revisit?

If you could know one thing earlier, what would it be?

Use the official links to explore food, wildlife monitoring, weather forecasting and ocean cleanup.

Remember one specific approach. Ask what else it needs to help in practice.

Open Market Marshes in Google AI Quests. This lesson uses an eight-minute discussion about data and testing.

What information might help predict rising water? How would you test a model with something it has not seen?

Pangu, GenCast, Aurora and Earth-2 differ in their tasks, outputs and evaluation conditions.

Check what was actually measured instead of comparing a single headline number.

A set of possible forecasts helps describe uncertainty.

The fraction of scenarios crossing a threshold is not an absolute guarantee for a place.

Observation, modelling, communication, reception and action can each fail.

A warning still needs people, routes, supplies and protection behind it.

Computer vision can produce false alarms and miss things.

Record lighting, distance, obstruction and sample conditions. Then review the results.

Wildlife monitoring connects equipment, recognition, expert review and conservation action.

A detection matters when it helps someone do useful work.

3D measurement records shape and position in space.

Resolution, obstruction and measurement error still matter.

Intelligent manufacturing connects sensing, planning and execution.

Measure errors, verify movements and provide suitable safeguards.

A solar power tower uses heliostat mirrors to concentrate sunlight.

Trace the energy, then distinguish solar thermal storage from nuclear reactors.

Collect heat during the day and store part of that energy in a thermal storage medium.

Operating time at night depends on storage capacity, weather and operating conditions.

Solar thermal salt storage does not involve nuclear fission. A liquid-fuel molten-salt reactor uses nuclear reactions.

The Fukushima Daiichi accident units were boiling-water reactors. These three systems should not be confused.

Decay heat remains after the nuclear chain reaction stops and must be removed reliably.

Cooling, pressure and passive provisions differ by design. Their capability needs verification.

Compare normal operation, accident releases and waste management separately.

Low-pressure designs may reduce some accident pathways. Materials, containment, monitoring and waste still need attention.

Effects can include radiation exposure, cleanup, relocation, livelihoods and long-term monitoring.

New designs need tests, regulatory review and operating evidence.

Research into microbial fermentation and starch made from carbon dioxide explores new production pathways.

Moving from an experiment to affordable, sustainable supply takes engineering and evaluation.

Look away from the screen and move around.

Then return to narrow your question and prepare a proposal.

Follow a young person's question about ocean plastic through design, testing and deployment.

Listen to his own account and check how the project records progress and limitations.

Before designing for someone, learn how they live and work.

Listen to the real difficulty. Respect anyone who does not want to be interviewed or recorded.

Pick a specific place, a group of users and an observable change.

Start where you can gain access and investigate.

Who experiences what difficulty, and under which conditions?

Which part would our attempt change?

Compare before and after, or use a suitable control.

Keep other conditions as consistent as possible. Record failures and surprises.

Restate their question before offering a specific challenge.

Check measurement, comparison, participants and real-world conditions.

Say who it is for, what you would try, how you would compare results and what remains uncertain.

Keep the main points on one worksheet.

Four groups each have three minutes to present and one minute for feedback: sixteen minutes in total.

As you listen, help the group spot a question they missed.

Choose one comment to improve the question, method or comparison.

Explain why you used it, or why you are keeping your original idea for now.

Write down one action for the coming week and how you will record its result.

Find someone willing to hear what you learn, and keep investigating.