AI Consulting

Hello Freddie. Could you introduce yourself and what you do at Aya Data and what you specialize in?

I founded Aya with our COO, Ama, in order to create an ethical data collection and annotation service – which we saw a huge need for.

We quickly developed a specialism in Geospatial annotation – and soon realised that we should also be adding a geospatial data science team to our business so that we could deploy our own custom AI solutions based on client needs. 

Today we have a number of geospatial AI products – using drones and machine learning to completely transform efficiency in farming, forestry, and infrastructure projects. 

We’ve built drone and AI solutions that can accurately count trees, predict crop yield, assess tree and plant nutrient deficiency and health, as well as solutions to measure the exact dimensions and condition of buildings. 

Since the launch of ChatGPT in November of 2022, the general impression is that AI is finding a home in almost all branches of industry. But AI with industrial applications had not started with ChatGPT. So, could you tell what effects the popularization of AI has had on your niche?

Mckinsey estimated recently that the contribution of Generative AI to the global economy will be up to $4.4trn annually.1 That’s like adding another Japan into global GDP, with a single piece of technology! The advent of ChatGPT was the sea change where most people really started to understand how significant AI will be

For us it’s meant a few things, we can annotate large data sets more quickly, we can deploy very powerful natural language models for clients more cheaply, we can run our operations more efficiently  – but the most major change has been awareness amongst our clients and potential clients of what’s possible with AI. Executives are more eager than ever to explore how AI can benefit their business.

And what effects has it had on geospatial AI in particular

Aside from stimulating more investment into AI generally – GPT / LLMs act as an amplifier to people and technologies. So now, if you have a geospatial analytics platform with an LLM interface, you can get far more value out of the interaction with the system. 

Imagine one day being able to ask Google Earth to quantify deforestation in the Amazon, or how much sorghum is planted this year in West Africa. Of course, even if a GIS platform knows what you want, it has to be able to do it accurately.

Relatively big news in the world of geospatial AI is NASA’s partnership with IBM Research to develop an HLS Geospatial Foundation Model – an open-source foundation model for Earth observation data. Could you tell us a bit more about that?

This is a great step towards what I mentioned before – a model that’s pre-trained to analyse satellite data in lots of different ways. I think they’ve had great success mapping flooding and other major geological events. 

The limitations here are that the data is from Sentinel 2 satellites and is accurate to 30m. This means that it’s a fantastic tool for analysing major changes, but not precise enough to be extremely useful for things like farming. 

What is the importance of foundation models for geoAI?

As we get bigger, better-trained foundation models that run on more accurate satellite data – the possibilities for generating insights on everything from climate change to farming practices are endless. The best part of all is that much of this is being made open source, for anyone to use and build on – so these insights will be democratized and verifiable by anyone.

What effect do you see the HLS model having on the services you provide?

Very little – we are focused on precision insights, that’s why we typically use drones to collect highly accurate RGB, LiDAR, and Hyperspectral data. We are keenly interested in watching this space though, and are looking forward to the day when we can develop the right level of insight from satellite data – particularly as this means we can add more value to smallholder farmers, where collecting drone data is currently financially unviable.

OK, so you’ve been very clear that the HLS model will likely have minimal practical impact on your specific line of work right now. Do you know of any emerging tech or practices that will?  

There is a huge amount of innovation happening right now in ML data pipeline software. If you look at companies like V7 and Labelbox, they started as annotation platforms, but have expanded rapidly to encompass the whole ML data pipeline, allowing you to ingest data, annotate it, train your models, and optimise through human feedback. 

These innovations make it far easier and faster for us to build and deploy accurate AI solutions at scale and we expect to see this trend continue.

Could you tell us what you believe will happen within the geospatial data science branch in the next 5 years? Do you see more breakthroughs, more widespread applications of geoAI, or maybe some developments that those of us outside of the industry can’t predict? What are your insights?

Analysts have the Geospatial market reaching $256bn by 2028, this is gigantic, but it wouldn’t surprise me. 2 As with all AI, the market size grows as quickly as we can think of new use cases, as well as taking leaps with technical breakthroughs. 

The range of applications stretches from carbon credit validation, to infrastructure planning, to disaster response and management. One thing is for certain, as data quality improves with better, cheaper satellites and drones, and stronger foundational models are developed, Geospatial AI will come further into the mainstream as an area of huge public interest. 

My favourite example of this recently was Narendra Modi committing thousands of agri-drones and flight training to women’s self-help groups across India, to help collect and analyse data and improve farming practices. 

For us the most exciting areas for geospatial are around climate and agriculture; how can we use drones, satellites, and AI to grow crops and forests more efficiently, profitably, and sustainably – with less fertiliser and inorganic inputs. These are the problems that we’re delighted to be working on at Aya.

Thank you for those insights Freddie and thank you for taking the time to talk to us.

Thank you for having me and I wish you all the best!

Sources:


1 McKinsey. (2023). The economic potential of generative AI: The next productivity frontier. [https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier

2 GlobeNewswire. (2023). Geospatial Analytics Market Worth $256 Billion by 2028 – Market Size, Share, Forecasts, & Trends Analysis Report with COVID-19 Impact by Meticulous Research®. [https://www.globenewswire.com/news-release/2021/09/16/2298690/0/en/Geospatial-Analytics-Market-Worth-256-Billion-by-2028-Market-Size-Share-Forecasts-Trends-Analysis-Report-with-COVID-19-Impact-by-Meticulous-Research.html]

Subscribe to Our Newsletter!

We don’t spam! Read our privacy policy for more info.

What Data do You Need to Create the Best AI Chatbots?

A CCO’s Perspective: A Chat About the Status of AI Right Now and AI in the Future