Today I wanted to write about a fun and certainly not dystopian topic: robots that build themselves. Surprisingly, this market remains largely untapped. While many factories feature a plethora of automation, it becomes increasingly difficult to eliminate humans in production processes when factors like gathering resources, maintaining the machinery, and shipping the products come into play. However, that doesn’t mean it hasn’t been achieved on a smaller scale.
The University of Cambridge developed a robot “mother” that builds “offspring” which would be subject to a series of movement-related tasks. The mother had a few “genes” to choose from when constructing the shape and movement constraints of the offspring. Then, the offspring that performed the best survived to the next generation while the offspring that performed poorly were subject to new genes based on mutations and the genes of the more successful offspring. Ultimately, this unnatural experiment about natural selection found that the mother was creative enough to build offspring that could double their original speed after multiple generations. Robot mama is still a bit large for an example of self-building robots. One of the most promising examples is MUCH smaller.
In November 2021, Harvard University’s Wyss Institute, a biological engineering group, published an experiment in which they introduced the world to Xenobots. Despite the Xenobot’s dreadfully cliché sci-fi name, they are almost unbelievable in what they are capable of. Researchers at the Wyss Institute took cells from the African clawed frog and had artificial intelligence develop an arrangement for these few thousand frog cells to allow them to gather other cells and reproduce. While that’s an extremely short explanation, they found that these groups of frog cells were able to gather other groups of cells and make “offspring.”
That reproductive process of gathering cells in an environment to make offspring is unique to Xenobots according to the study, but how does this connect to robots making robots? Artificial intelligence designed the structure of the Xenobots. In contrast with the robot mother who built its own offspring, this AI ran virtual simulations to come up with a structure for Xenobots to allow them to rapidly create offspring. The AI produced a structure reminiscent of Pac-man.
Both these examples illustrate two separate ideas of using robots to make more robots. Cambridge’s robot is a more literal and real-world example of a robot using natural selection to make better robots, but Harvard’s Xenobots are an example of using AI to better optimize a more natural reproduction process. I think Harvard’s study is much more promising because they plan to continue to use Xenobots for medical procedures where they use the computer-generated structures for specific tasks like fighting cancer.
On the other hand, Cambridge’s study certainly enters the gray area surrounding complete autonomy by allowing a robot to continually make better versions of itself. That too could have some useful impacts for optimization, but I think I’d sleep better at night knowing that there isn’t a factory where robots constantly build stronger and stronger robots that can move faster and faster, but maybe that just makes me a technophobe.
Be First to Comment