Skip to main content
ProjectBeginner45 minutes

AI-Powered Research: The Unmapped Ocean

Research a single topic across four AI models and synthesize their answers into something better than any one model produced alone.

What you'll build

A short, well-sourced research brief on one of the most surprising facts in science: we have better maps of Mars than of Earth's own ocean floor.

What you need

Steps

  1. Ask your first research question

    ~5 minutes

    Open any one of the four AI tools and paste the prompt below. You'll get a clear, structured answer covering key facts. Save the response in a text file and label which AI you used. This is your baseline — a decent answer from a single source.

    Prompt
    I just learned that we have mapped more of the surface of Mars than Earth's ocean floor. Is this actually true? Give me the key facts: what percentage of the ocean floor is mapped, what percentage of Mars, why the difference exists, and what we might be missing down there. Keep it under 400 words and cite specific programs or missions where you can.

    What to expect

    A clear, structured answer mentioning sonar limitations, the Seabed 2030 project, and a comparison of ocean mapping costs to space missions.

    Try this

    Ask the same question on a different AI after completing the full project and compare how the baseline answers differ.

  2. The Science Angle — Gemini

    ~5 minutes

    Open Gemini and paste the prompt below to pull out the technical and physics perspective. Save the response and label it: "Gemini — Science/Technology angle".

    Prompt
    Explain the technology behind deep ocean mapping. What are the physical limitations of sonar versus satellite radar? Why can we map Mars from orbit with cameras but can't do the same for Earth's ocean floor? I want to understand the physics — why water makes this so hard. Include specific numbers: depth, resolution, speed of sound in water vs radio waves in space.

    What to expect

    A technically detailed answer about acoustic mapping, multibeam sonar, and why electromagnetic waves don't penetrate seawater. Gemini tends to be strong on scientific detail.

    Try this

    Follow up: "What is the current best resolution achievable with multibeam sonar at 6,000m depth?"

  3. The Exploration Angle — ChatGPT

    ~5 minutes

    Open ChatGPT and paste the prompt below to pull out narrative and discovery stories. Save the response and label it: "ChatGPT — Discovery/Exploration angle".

    Prompt
    Tell me about the most significant discoveries made on the ocean floor in the last 20 years — things we didn't know were there until we looked. I'm interested in geological features, ecosystems, shipwrecks, or anything that surprised scientists. What might we still be missing in the 80%+ of the ocean floor we haven't mapped in detail?

    What to expect

    Stories about hydrothermal vents, species discovered in the deep, and underwater volcanoes taller than mountains on land. ChatGPT tends to be strong on narrative and examples.

    Try this

    Ask: "What deep-sea discovery has had the biggest impact on our understanding of how life began on Earth?"

  4. The Money and Politics Angle — Grok

    ~5 minutes

    Open Grok and paste the prompt below to get the economic and geopolitical perspective. Save the response and label it: "Grok — Economics/Politics angle".

    Prompt
    Why don't we fund ocean floor mapping the way we fund space exploration? Compare budgets: what does NASA spend annually versus what ocean mapping programs like Seabed 2030 or NOAA receive? Who benefits from better ocean maps — military, shipping, mining, climate science? Is there a geopolitical or economic reason the ocean floor stays unmapped?

    What to expect

    A direct, sometimes contrarian take on funding priorities, military interests in bathymetric data, and deep-sea mining politics. Grok tends to be blunt about economic and political angles.

    Try this

    Follow up: "Which nations have the most strategic interest in keeping ocean floor data classified?"

  5. The Human Story — Claude

    ~5 minutes

    Open Claude and paste the prompt below to get the human and organizational perspective. Save the response and label it: "Claude — Human story angle".

    Prompt
    Who are the people trying to map the entire ocean floor? Tell me about the Seabed 2030 project and any other major efforts. What drives someone to dedicate their career to this? What's the current status — how much progress have they made since starting? And what does "mapped" actually mean — what resolution counts?

    What to expect

    A thoughtful, nuanced answer about the Nippon Foundation-GEBCO partnership, the definition debates around resolution standards, and progress updates.

    Try this

    Ask: "Who is the most important individual contributor to ocean floor mapping in the last decade?"

  6. Compare what you got

    ~5 minutes

    Read through your four saved responses and write quick comparison notes. Identify facts that all four agree on (likely reliable), claims that appeared in only one response, and any contradictions between models — especially differing numbers, dates, or percentages.

    What to expect

    You will find at least one factual discrepancy between models. This is the point — you have found something worth noting in your final brief as an honest range of estimates.

    Try this

    Rate each response 1–5 on depth, accuracy (based on what you now know), and readability.

  7. Synthesize with AI

    ~10 minutes

    Pick any one of the four models. Paste all four responses into it along with the synthesis prompt below — replacing the placeholder sections with your actual saved responses. The AI will catch contradictions you missed and produce a research brief that draws on the strongest parts of each.

    Prompt
    I researched ocean floor mapping using four different AI models. Below are their responses, each from a different angle. Please:

1. Identify the 5 most important facts that multiple sources agree on
2. Flag any contradictions between the responses (different numbers, conflicting claims)
3. Note the most interesting insight that only appeared in one response
4. Write a 300-word research brief that synthesizes the best of all four into a coherent summary
5. List 3 questions I should investigate further if I want to go deeper

Here are the four responses:

--- RESPONSE 1 (Gemini — Science) ---
[paste Gemini response here]

--- RESPONSE 2 (ChatGPT — Exploration) ---
[paste ChatGPT response here]

--- RESPONSE 3 (Grok — Economics) ---
[paste Grok response here]

--- RESPONSE 4 (Claude — Human Story) ---
[paste Claude response here]

    What to expect

    A synthesis genuinely better than any single response. The AI will catch contradictions you missed, surface patterns across all four, and produce a research brief drawing on the strongest parts of each.

    Try this

    Ask the AI to write a second version of the brief aimed at a non-expert audience — compare the two versions.

What you learned

  • How to use four different AI models for free
  • How to frame the same question from different angles (science, narrative, economics, human)
  • How to spot where AI models agree and disagree
  • How to use AI to synthesize multiple sources into a coherent brief
  • Why single-source research (one model, one prompt) is insufficient

Go further

Harder: Pick a controversial topic (AI regulation, nuclear energy, gene editing) and run the same multi-model method. Notice how the models' biases become more visible on contested topics.

Different angle: Try the same topic but ask each model to argue a position — one argues for more ocean funding, one argues space is more important. Then synthesize.

Icelandic connection: Iceland sits on the Mid-Atlantic Ridge — one of the most geologically active underwater features on Earth. Ask any model: "What role does the Mid-Atlantic Ridge near Iceland play in ocean floor mapping research?"

Troubleshooting