Podcast: AI in the built environment.


The design flywheel and the fly in the ointment

In this episode, our host Ellie Griffiths chats to Alan Mosca, co-founder and chief technology officer at nPlan – a machine learning-based platform that forecasts and predicts outcomes of construction, engineering and infrastructure projects.

Imagine you’re designing a building and you have 100 designers with you. You can create 100 options, which each one of them makes, then you select the best one and have them make 100 versions of that.

As well as the capability for innovation through optioneering, improving efficiency, and the flywheel that AI can give to designers, we discuss sustainability and energy consumption cost-benefit analysis, efficient AI chips of the future, and cautionary tales.

You are always taking two steps forward and one step back. This is true of all step-change progress: the internet, the telephone, the car. You always have a social adjustment. It gets worse before it gets better.

Alan offers his perspective on concerns about bias in data sets, plus a new bias that he believes is soon to come, arising from models trained on the output of other models, and the risk of this ‘hallucinatory’ knowledge being perpetuated.

Take a model that helps you with engineering: what if it gets a calculation wrong? Even though self-driving cars are less likely to make a mistake than a human, who’s responsible for the mistakes? There are problems that need to be untangled.

Find our latest chat on iTunes by searching ‘Hoare Lea’, or listen below.