Banner by Anthony Pascarella

AI's Worst Fear: A Dirty Kitchen Sink

by Anthony Pascarella

I promise you that robots will not be enslaving humankind anytime soon. You can sleep easy, at least for tonight. Artificial intelligence has proven to be a powerful ally in countless areas where our squishy, organic compositions are not a good fit. The cold and calculating logic of computers far surpasses our own abilities in many areas of life. Today’s world runs on machinery, after all.

While there is no denying the superior processing power of computers, this does not mean that our circuit-brained counterparts are infallible. The greatest weakness of artificial intelligence stems from its greatest strength: logic itself. A computer’s ability to systematically solve a problem becomes a hindrance when faced with a puzzle that incorporates multiple, dynamic steps. AI today lacks the very thing its creators used to build it in the first place: creativity and free thought.

As it turns out, these skills are rather important in many everyday tasks that we barely pay attention to. These menial jobs are almost second nature to us, so much so that we often have time to complain about how bored we are while doing them. While it is easy for us to do household chores, artificial intelligence still struggles with many of these undertakings. It might surprise you to know that robots still cannot do the dishes. And by doing the dishes, I do not mean simply spraying them down with soap and rinsing them like your average dishwasher. I am talking about the kind of chore where you come home from a long day ready to relax, only to be met with a sink piled high with dirty plates and utensils.

Now, before you discredit the dishwasher you may have at home, it is important to know that it was never meant to be the final solution to washing the dishes. Yes, it cleans off all the food stains and sanitizes your kitchenware, but someone still needs to load, unload and organize the dishes (yet another chore reviled by nearly everyone). Recently, a startup company named Dishcraft rolled out a massive, industrial-sized dishwashing robot aimed at restaurants and cafeterias. It may sound great in theory, but this robot is nothing more than a glorified, automated dishwasher. Though it pulls dirty plates from one cart and deposits clean plates on another, a human employee still must wheel these carts around. The robot also only washes a certain type of plate outfitted with a small metal chip on the bottom, since it uses magnets to hold the dishes while they go through the cleaning process.

Dishcraft’s invention is certainly a step in the right direction, but there is still a long road ahead for robotics and artificial intelligence in order to tackle the age-old chore in a way that could benefit the everyday household. The issue? Doing the dishes is not as simple a task as you might believe.

When you really break it down, there are numerous small actions and decisions that make up each step we carry out automatically. Not only do we need to pick up and rinse the dishes off, but we also subconsciously calculate how many plates of various shapes and sizes we can stack in the sink or on the neighboring counter. We can theoretically predict what will happen when we put various items under the faucet and move them around. (I still struggle with this concept when it comes to spoons; I always manage to spray myself with water no matter how hard I try to avoid it). While this is happening, we also mentally map how we are going to organize the dishwasher or drying rack to fit all the dirty dishes. Put all these little pieces together and suddenly, a little chore becomes quite a complex undertaking. 

“That turns out to be an extraordinarily difficult problem,” says Google’s former CEO Eric Schmidt. And we, as humans, solve this puzzle in a matter of seconds. We utilize years of accumulated knowledge of the physical world in harmony with practiced movements honed by muscle memory to carry out these sorts of chores with little acknowledgement at all.

AI, however, lacks this innate ability to reason or use context when faced with such an undertaking. Computers excel when faced with very small or specific challenges that can be solved through a series of predefined steps and protocols. This is why most of the robots built to date are meant for a very select purpose, such as building a certain part of a car or carrying out a specific type of analysis. The smaller a robot’s environment is, the easier it becomes for it to analyze the problems it faces and calculate a viable solution to them. The problem with cleaning dishes is that the potential environment is so large it is nearly impossible for our current technology to account for all the tiny details that a robot will face.

A team of scientists have recently developed a solution to this issue. Researchers at MIT are training computer software to do the dishes by feeding it numerous videos and simulations of objects being moved around. Advancements in machine learning now allow computers to “learn” in a similar fashion as humans do. The MIT program is now able to recognize patterns in the way physical objects react to various touches and motions. Over time, the computer will be able to predict these reactions on its own. Once it reaches this point, robots can be designed to utilize these newfound contexts to accurately pick up, clean and stack dirty dishes and utensils.

When will we start to see robots pop up in our kitchens? This won’t be happening anytime soon; especially not in the sense we may expect. But innovations like the Dishcraft system are certainly a step forward in the world of household robotics. We already have the iRobot Roomba, after all.

For the time being, at least, it looks like we are stuck doing the chores we always hate to do. Perhaps that is a good thing. After all, doing the dishes is quite a complicated task. The next time you find yourself facing a sink piled high with food-stained dishes, just know that you are about to embark on a mission our robotic counterparts cannot even imagine.