Huawei wins SLA's OneMap GPT Challenge

The Huawei OneMap GPT is in prototype - and it can already do things that Google Maps cannot.

Visited Geo Connect Asia this morning. It was my first time there and I learned a great deal about the possibilities of geospatial.

Created by a team at Huawei, OneMap GPT adds generative AI to SLA's OneMap - and is one of two winners of the OneMap GPT Challenge.

I was very impressed by what it could do.

🚀 There's a GPT for that

OneMap GPT looks like Google Maps but with a chat interface in place of a search box.

The service was demonstrated on stage, where it was prompted to generate:

- A list of homes based on budget, proximity to MRTs.
- Check if address is within 4km of parent's home.
- Calculate housing subsidies that one qualifies for.

It could do other things too, such as suggest travel routes across Singapore, places to visit, and available cars for rent using text prompts.

And this is just the prototype.

📈 Graph database inside

Had the pleasure of meeting Ashley Fernandez, who described how his team created it.

- Prototype builds on initial R&D by his team at Huawei.
- Data is currently static, but integration's an API away.
- A graph database was used to define relationships.

Ironically, the biggest challenges had nothing to do with generative AI, Ashley told me.

🔸The incorporation of written policies was challenging, as policies can be complex and are processed using natural language processing.

🔸OneMap GPT was only possible due to R&D work that opened the door to stitching SLA's geospatial data with multiple data sources using graph database tech.

🕰️ When will it be available?

There is no release date slated at the moment. However, it is worth noting that OneMap GPT was built in slightly over two months. By 5 people, including Ashley.

In fact, the current UI was only polished up over the last 2 weeks. I imagine the team is working hard to bring it to release.

For now, you can visit Day 2 of Geo Connect Asia on Thursday to catch it in action.