Get down. I think people hear "robot" and they think Terminator or Robocop. Your move, creep. The idea that you could build a new organism that has never existed from cells that does something we’ve never seen before, essentially a biological machine, has really captured the imagination. If xenobots sound like science fiction brought to life, it's because they are. They’re biological machines made by humans but designed by computers. This project offers a new way to make what you might call a robot that's very different from existing robotic technologies. 



A machine for a new purpose, not to recapitulate biology or build a heart, but to build something that could do work for any number of tasks that you'd use a traditional robot for. So it's kind of a weird marriage between robotics and computer science and biology. These tiny specks seen here under a microscope are cenobites, made of about 5,000 living skin cells. They get their name from the African clawed frog Xenopus Laevis, which supplies their cells. They can't reproduce and they will only live for around 10 days, feedingon the small platelets of yolk that fill each of their cells. 

Really, this was the first time we thought, could you take stem cells and build something from the ground up? And the idea is really to take these stem cells very early in development, collect them and essentially have different types of cells, like you would different Lego blocks. Cells already act like little machines, some contract and expand, others capture material. By shaping these cells, you create the machine. So you could imagine a red Lego block is a heart cell and a green Lego block is a skin cell. And then really similar to building Legos where you just sort of start connecting the pieces from the bottom up, we build our robots in the same way.

 We just build cell by cell until we get to a final shape and form. Xenobots are designed and then evolve using artificial intelligence before they're made. Sam Krieg man is a graduate student at the University of Vermont and designed that computer program. Unlike every other organism that has ever existed before on this planet, the xenobots evolve inside of computer simulations and the computer simulates the physical world, but it can do so very rapidly and many worlds at once and so, it can do evolution in days instead of millennia and at the end of this evolutionary process, it prints out a blueprint that tells us how to build a new organism for some desired task from essentially scratch from existing biological materials.

 The simulated xenobots are subjected to physics engines similar to those in video games. The AI assesses how variants will perform and makes changes according to the tasks they're set. This is natural selection put on fast-forward with a computer at the controls. Evolution on Earth works as a long chain of mutations, Evolution on Earth works as a long chain of mutations, tiny revisions to existing designs, one after the other over millions or billions of years. And sometimes, we can speed this up by selectively breeding crops and livestock and dogs so that their offspring have desired characteristics. But it's very unlikely that you could select from wolves or dogs anything that you wouldn't call a dog. But with computers, we can speed up this process.

                                                 

 So we can simulate billion years of evolution in a day or in a week on a super computer. The computer selects its top candidates to build and only then do humans come back into the picture. The AI generated blueprints are sent off to Dr. Blackstone and his wet lab at Tufts University, Massachusetts. Right, so this is the fun part, it's a bit like sculpting and some designs are not possible to be built biologically currently. They have features that are too difficult to shape or construct like right angles, very small gaps. 

But we often get five or six excellent designs that are completely buildable in the lab, which is where we port everything over into the wet lab and build them by hand manually. We then build from the ground up using those different cells the shape that new one, and it typically starts as a sphere. So the cells round off and make a roughly round shape after 24 hours and then we go in with a pair of surgical forceps that we polish with the sharpening stone manually and also a very small cautery electorate.

We carve away from that sphere the design and sculpt the final shape that we would like to match the computer design and that's really sort of the amazing part in the innovation is this, this sculpting aspect to give you a very, very precise three-dimensional shape at only the scale of about half of a millimeter. Right now, xenobots are made of only about 5,000 skin and heart muscle cells, but they can perform basic tasks like moving in a straight line, carrying objects from one spot to another and behaving collectively.

 Xenobots don't need to be that sophisticated to supply some kind of social utility. We found that even the xenobots that simply swirl around in their dish, they tended to clean their dish as they moved through it. So that is, they tended spontaneously to sweep up and corral any detritus in their dish into centralized, neat little piles. Why they do that is still somewhat of a mystery. We evolve them in the computer simulation simply to move and we got something for free because they're living systems that have their own agenda and do useful things that weren't necessarily programmed, like cleaning.

 At the moment, all xenobots are constructed out of frog cells, but if human cells were used instead, then there's huge medical potential for these machines. Potentially, they could simply by swirling around in an artery, scrape plaque off of the walls or abrade calcium deposits in arthritic joints. or abrade calcium deposits in arthritic joints. With a little bit more bio-engineering, we can imagine targeted drug delivery, which is something that we’ve been thinking about even in our very simple simulations. So, when we evolve inside of a computer a xenobot to carry a little pellet from one position to another target position, what that's getting at is, could these be used to transport something useful, not just a pellet, but a drug or medicine to the precise location of disease?

                                             

 And so, the advantage of using a human patient's own cells is that the immune response system wouldn't flush it out, unlike other micro robotic delivery systems made of metal, for example. If the production of xenobots, which exists right now, could be scaled up hugely, by automating them with 3D printers, then they could be used to tackle issues on a global scale, such as pollution. We could imagine without too much further bioengineering simply swarms of xenobots that swirl around and clean, simply swarms of xenobots that swirl around and clean, could be used in the environment to clean up little microscopic bits of plastic, to aggregate them into a larger ball of plastic that could be identified and picked up by traditional robots, boats, or drones, and then taken to a recycling center. Just wear and tear on robots is awful difficult.

 So, it's very easy to lose functionality with small changes, especially when you’re dealing with physical motors. It's fairly catastrophic when there’s any sort of damage. But biological systems for millions of years have been optimized to resist wear and tear and damage and certainly, amphibians have many of the same pathways that we humans do. So, leveraging that's been great, it gives us the ability to have organisms that when they're damaged, can continue functioning even with fairly catastrophic damage. 

So you can largely bisect the xenobot and it will zipper itself back up in about five minutes and continue on doing what it did before and this is really amazing. We haven't seen another system that can do something to that extreme. And Blackiston and Krieg man believe it's not just the xenobots that are revolutionary, but also the supercomputer that’s used to create them. One day, it could potentially help humans redesign themselves. We have supercomputers that can run millions of simulations per minute and it also can be optimized for whatever we would like. So as humans, we often have goals that may be slightly different than natural selection. 

And we can now have the computer set up evolution and sort of get through this bottleneck much faster. So for example, for the heart, we're seeing that with the types of diets that humans eat now, there may be some other type of heart valve or even some process in the body or vasculature that's superior to what we humans have and the computer can likely hone in on those types of changes very quickly. I think that the point of the computer simulations I think that the point of the computer simulations isn't just to speed up evolution.

                                           

 It's that human designers have really strong biases It's that human designers have really strong biases in the kinds of artifacts that we make. So, when a human goes to design a robot, they inevitably will copy some familiar object. The robot will look like a dog or a human or a plant that they've seen before, some kinds of shapes that are familiar to them. Now, the computer doesn’t share our biases. It might have it's own biases in it. 

But it will come up with designs that look nothing But it will come up with designs that look nothing like any organism or thing that we've seen in the world And then it's a matter of tackling all the big ethical and societal questions about should those be implemented? and societal questions about should those be implemented? How much control should we have over our bodies and the design of humans or other organisms? These are the same things that the plant scientists have been grappling with, right? It's can you make a bigger tomato? Can you make a bigger apple? Should you? Should you make it pesticide resistant? 

And I think what we're going see are those questions moving into humans very soon. We're starting small, we’re using xenobots, but you can extrapolate this on up. So, if a computer is so good at designing a microorganism, could the computer design a better horse? Could the computer design a better cat? Could this computer make something that's more efficient, a better heart for humans, if given the right amount of information and the right amount of synthetic evolution? And so, I think there's a real push to understand how AI can be harnessed to really capitalize and even push where evolution has gotten us to this point.