The recent sale of 43 global ports by Li Ka-shing—including two along the Panama Canal—has drawn international attention. I attempted to develop an interactive Cesium visualization showing these ports' geographic distribution, but news reports only mentioned quantities without specific coordinates. This required manual data collection.


PS: Video demonstration available on MalaGIS Video Channel.

Initial Approach

My plan seemed straightforward:

  1. Find the official announcement with a port list
  2. Use Google Maps API to geocode port names into coordinates
  3. Compile results into a CSV

I even prepared Python scripts for batch geocoding. However, news sources focused on geopolitical implications rather than technical details.

Grok AI's Breakthrough Methodology

As a last resort, I turned to X (Twitter)'s integrated Grok AI. Its solution exceeded expectations. Observe the workflow:

Grok's systematic approach:

  1. Translated query to English
  2. Located Bloomberg article via search
  3. Identified missing port list in article, traced to CK Hutchison's press release
  4. Corrected broken link via "CK Hutchison Panama Canal ports" search
  5. Analyzed operational ports on hutchisonports.com
  6. Excluded Chinese ports (e.g., Shanghai, Ningbo)
  7. Supplemented gaps using CK Hutchison's annual report PDF
  8. Cleaned and validated data

It produced 39 ports—short of the reported 43, but with meticulous validation surpassing manual efforts.

DeepSeek Performance Comparison

Tested Tencent's Yuanbao (DeepSeek-based):

Even with refined prompts:

Result: Only 12 ports identified—likely due to Chinese-language bias in training data.

Data Accuracy Notes & Download

Grok's output contained errors (e.g., missing Sydney's southern latitude), likely from web searches rather than geocoding APIs. The curated dataset remains valuable for experimentation:

Join MalaGIS Telegram Group to get more infomation.

Conclusion

Grok's structured problem-solving demonstrates AI's evolving role:
Past: "Make AI do my work"
Present: "Learn methodologies from AI"

For GIS professionals, this signals a paradigm shift—embrace AI not just as a tool, but as a collaborative strategist for spatial data challenges.