There's various ways. One really popular way at the moment is called biomimicry, which is where we look to nature, for example, wool, a natural material that has loads of properties that we want. And we try and work out how does it do it? What's it made of? And the brain of a material scientist is about zooming all the way into the atoms and seeing what type of atoms they are, how they're bonded together. And then zooming out just a teeny tiny bit and then looking at the molecules of those materials. How are they structured together? Are they in a very, very neat array, or are they all jumbled up together? Are they bonded stiffly together or is it a little bit more loose? Zooming out a little bit further and seeing if there are any textures in those groups of atoms and then zooming out entirely and saying, okay, so with the knowledge of how this material is built from the very tiniest

of scale, what can that tell me now about why this is strong or tough or flexible or biodegradable? Evolution by natural selection produces these remarkable things. But I know we were talking earlier, Mark, about the idea that you can use that sort of syntactic evolution to have the materials designed themselves. So,

yeah, we have these things called 3D printers now and we can basically layer materials and make new materials by layering them in different combinations. And so a lot of the time we use theories to make the material we want to make and we have a particular idea of it. But a different way to do it is to just have a computer control that 3D printer and say, okay, you've got these 10 materials you can combine in any way now. Or if you go combining them, if one of those materials can do some function which we're going to test it on better than the next one, we'll reward the algorithm so we use a sort of artificial intelligence algorithm to keep feeding back into the system. As these materials get produced and tested, produced and tested, produced and tested, it's constantly saying, okay, that worked. I'll do a bit more of that. That worked. I'll do a bit more of that. And if you keep doing that, you have artificial evolution on a 3D printer. And potentially you can sample a whole load of material property space that you might not actually have thought were possible and discover something

completely novel.