AI in Biodiversity Net Gain + Better Batteries
Here we are exploring recent advances in the use of AI in material science and biodiversity. What can the real estate industry learn from them?
This was originally produced as a podcast, you can listen here. But for those of you who prefer to read, I hope you enjoy the below.
NEW, BETTER BATTERIES
At the end of last year we spoke about Deep Mind’s research into materials, they have been able to predict 2.2 million new potential materials for technologies such as super conductors and batteries (link).
Out of those 2.2 million materials 380,000 are believed to be the most stable and potentially useable. For context, over the last decade 28,000 new materials have been discovered. Scientists do this by tweaking known crystals or experimenting with new combinations of elements. This is an expensive, trial-and-error process, taking months to deliver even limited results.
This example shows we can use AI to undertake some parts of this trial and error process. The results moved us from 28,000 new materials in a decade to 380,000 last year alone.
It’s not just Google & Deep Mind who are doing this.
IBM have a material discovery AI called RoboRXN, researchers at Stanford and MIT are working on it and Microsoft earlier in January 2024 announced they have found a promising new battery material (link). It looks like it’s less likely to burst into flames than today’s lithium-ion batteries. Importantly, it also uses less lithium, which is getting harder to come by as demand soars for rechargeable electrical vehicle batteries.
Microsoft used their AI called Azure Quantum Elements which helps researchers simulate experiments. They were able to find 32 million different candidate materials, filtered this down to 500,000 most stable materials and then filtered those down looking at how easily they could conduct electricity and how practical it would be to create those materials. This left them with 23 potential new materials.
All the whittling down took just 80 hours. Obviously, there’s a huge amount further R&D to develop the materials and test them in the real world. However, this initial process of trial and error is being sped up by such magnitudes, it should materially improve the speed and confidence with which new, better, more efficient materials are created.
EFFICIENCIES IN RAW MATERIAL MINING
On this topic of raw materials of lithium there is an interesting corollary with AI in the mining world.
According to The Economist 99% of exploration projects looking for new sources of mineral deposits fail (link). 99%. A great proportion of the cost of raw materials is therefore on the R&D that goes into failed search projects. There are now a number of companies who are training AI models on geological, geochemical and geophysical data to decrease that 99% failure rate.
DUAL EFFORT TO INCREASE ENERGY EFFICIENCY
These two advances create pressure to improve the types of materials we have on two sides.
On one side AI is helping to reduce the quantity of materials needed to help us with our energy needs. On the other side AI is making it easier for us to find the materials that we do need.
All of this makes me feel much more hopeful about the energy transition needed.
HOW ARE NEW MATERIALS RELEVANT TO REAL ESTATE?
Firstly, directly, the more, the better materials we have for batteries should make it easier, much more cost effective for us to transition to powering our buildings with cleaner energy.
Secondly, what about the other materials we use in our industry?
I had an interesting conversation with a brainy engineer recently who believes that there are better plasterboard solutions than Gypsum but that in our risk adverse, and particularly fire conscious world, it remains the easy, safe bet. This is despite the major potential issues with disposal down the line.
So could we bring these new material testing techniques to our industry? Help us find more sustainable, better insulating materials that are safe to dispose of or easier to recycle? At the moment, frankly I expect the costs are too high to use these research tools. Additionally, our purchasing market is too fragmented. Perhaps it’s something the big contractor-developers the Lendleases, the Skanskas have the right R&D budget for. I’ll keep my ears and eyes peeled and let you know if I hear of or see anything like it.
AI IN BIODIVERSITY NET GAIN
So that’s AI in materials research, let’s take a little left turn into AI in biodiversity.
This may sound off-the-wall but bear with me. I think there’s great potential relevance to real estate
One way of tracking biodiversity is to listen to animal calls, get experts to identify the calls of different animals and then work out how much wildlife there is. Of course this only identifies the animals who actually make noise but the research proved that it’s’ a good proxy for wildlife in general.
This is a similar approach to field studies for your ecologist reports when a surveyor goes out and counts the species in a field or in a set square.
Last year a group of German and Ecuadorian scientists wanted to measure biodiversity recovery in forests, farmlands (link)
In this Ecuador project, they trained a machine to listen to sound recordings of the forests they were looking at and identify things like bird calls, insect noises. In their tests they found that the AI could identify bird calls and insect noises as well as the human experts.
Imagine, this could potentially give us an automated, standardised and reliable way to report on biodiversity.
SO WHAT IS THE IMPLICATION FOR THE REAL ESTATE WORLD?
This research is interesting for the real estate world from two fronts firstly most practically. Obviously from this year, 2024, the Environment Act requires a 10% biodiversity net gain for new developments.
Currently this is based on a field survey and existing landscape plans and so on. These are respectable methods and probably the best we have right now.
NEW METHODS FOR BNG CREDITS
But just imagine if you could have an acoustic monitor on your site consistently measuring the levels of biodiversity. It would give us much more confidence that we are actually improving biodiversity. Think about the market for Biodiversity Net Gain credits. An approach like this could increase the robustness of the biodiversity metric tools, developing a stronger market incentive to increase biodiversity.
This sounds similar to carbon offsets and there are a number of potential issues and unintended consequences. This is not a panacea but it’s a good example of AI not just improving existing systems but creating new, more reliable ways of doing things.
THE DEVELOPERS REPORTING REQUIREMENTS
In the context of development, over the last 5/10 years the amount of reporting for planning applications has increased so much for developers.
Now to be clear, these increased reporting requirements are generally for the good and produce better, more sustainable buildings, which are better for people.
Producing ecologist reports, active travel plans, nutrient neutrality reports and so on lead to better buildings.
However, they do also make our planning process slower and more expensive. Quite simply the slower and more expensive the planning system, the more expensive homes become, the less affordable housing developers say they can afford to deliver and so on.
Anything which can speed things up planning and more importantly, more reliably show the environmental & ecological performance of development, the better.
In my experience, at the moment, many developers (not all) but still many, see these reporting requirements as a tick box, just another report they have to submit with their planning application. The sooner that we can create tools and methods to show it’s not just a regulatory tick box but a transparent and trusted method to judge a developments ecological success the more developers will start to use it as a point of differentiation and see it as a true market advantage.
Practically, the tools are not in place to do this yet but these examples show the potential. With so much pressure to reform the planning system, my bet is this is a 5-year horizon type technology not 50 year.