Learning about AI has led me to believe it could be transformational for the real estate industry. Generally AI is better at interpreting complex and disparate data than traditional computing. Goodness knows there’s a lot of complex and disparate data involved in real estate! In this article I explore this idea a little more and raise some of my concerns.
Programming Computers for Real Estate
Real estate has been notoriously difficult to digitise.
Each asset is physically & legally unique. Buildings also change over time.
These changes happen naturally due to movements in the ground or general wear & tear of building materials. Or changes happen when we refurbish and extend. The financial position of a building changes when we lease it. The legal position might change due to external policy changes.
All of this makes it difficult to program computers, using traditional computing, to accurately analyse all the different physical, legal and financial elements of the building.
Traditional Computing vs AI
To think about the potential of AI with real estate, we need to understand the difference between AI and traditional computing.
Traditional Computing
Our current computers do whatever we tell them to do. They are like electronic factories. We put data in, it performs actions on that data (actions that we’ve told it to do) and it produces the requested result.
For example, to make the search function work in our phone’s contact books, a programmer would have written a code that says something like:
“IF letters typed in the search bar EQUAL a name in the phone book, then SHOW the associated telephone number”.
This is one of the reasons we get glitches on computers. A glitch is basically a scenario that a human has not accounted for in the code for a program. When app developers talk about “bugs” they are really talking about a scenario they hadn’t considered, so the computer doesn’t know how to deal with it.
As Mo Gawdat says:
Q// Throughout our short history of building computers, we have always been fully in charge. The machines obeyed our every order. Every instruction, contained in every single line of code, has always been enacted exactly as we determined…They did exactly what we asked them to do, nothing more.//
(Scary Smart, Gawdat, 2021)
I believe this is one of the reasons it’s so hard to digitise real estate processes.
We would have to program every single possible outcome into a machine. As each building and transaction is unique, the range of possible outcomes is too great for any coder to teach a computer.
How is AI Different?
AI differs because computers are taught to make their own conclusions.
With AI, computers are given sets of data and programmers ask the AI to find patterns or clusters of data.
As an example, Google’s Deep Mind (their AI business) used AI to reduce energy usage in their data centres by 40%.
The Deep Mind team say that it is hard to use traditional rules-based formulas to optimise energy usage in data centres because:
“The equipment, how we operate that equipment and the environment, interact with each other in complex, nonlinear ways.”
Deep Mind 2016 link
To train their models, Deep Mind took the historical data that had already been collected by thousands of sensors within the data centre – data such as temperatures, power, pump speeds, etc. – and used it to train deep neural networks (more on neural networks here and deep learning here).
The deep neural network basically means the computer trains itself to find connections without a programmer having to tell it things like “adjust the temperature to X inside then data center when the weather is expected to be Y outside the data centre”. The computer teaches itself how it should adjust in these scenarios to minimise energy used.
The Big But
But this example leaves one big question, around data…
Deep Mind was able to achieve their results because they had a wealth of data to train their AI models on.
At the moment, the amount of data being collected about real estate is inconsistent. Therefore I think it may be hard to achieve similar results in other scenarios.
There are many companies working on improving data collection in real estate and these will make vital contributions.
Additionally, we may be able to harness other untapped sources of data. For example, the agriculture industry is starting to use satellite imagery to analyse crop performance. In real estate, perhaps satellite heat imaging could be used to assess the energy performance of new developments.
So, What’s Next?
In short, I don’t have the answers yet for how we could use AI in real estate.
However, the complexity of real estate makes me think there is real potential for AI to provide invaluable insights into the industry.
Therefore, I will keep learning about AI and sharing thoughts on how we could practically harness it for the good of the industry!
People | Spaces | Places 