In partnership with

Back in March, this post went viral on X (Twitter):

Then it went viral on Instagram. And then Linkedin. And then Instagram, again.

The post's author, Joanna Maciejewska, later clarified that she didn’t mean actual laundry-folding robots but rather for AI to handle tasks we hate instead of those we enjoy.

If, like me, you would very much like literal laundry-folding robots, keep reading to find out if or when we can expect that reality 🤖 

🌿 Let’s grow!

Today’s ammi.ai is brought to you by:

The Daily Newsletter for Intellectually Curious Readers

• We scour 100+ sources daily

• Read by CEOs, scientists, business owners and more

• 3.5 million subscribers

Sign up today!

Planting the Seed: explore AI topics and headlines in simple language.

Is AI on its way to do your chores?

The State of AI Household Assistance Today

AI and robotic assistance are already present in many of our homes.

Perhaps you have appliances and devices in your home that are connected to the internet, so you can command a smart home speaker to turn all the bedroom lights off or play a specific movie on your living room TV.

Robotic vacuums do a decent job of diligently scooting around your home, picking up dust and debris.

However, many of the most tedious household tasks, like cooking, cleaning, and folding laundry, are also the most physical and draining. And these tasks mostly remain beyond AI-powered robots’ current capabilities.

Why is that?

Sensorimotor Woes

Last summer, I had the opportunity to attend a presentation by the former Head of Facebook AI, Dr. Jitendra Malik, called 'The Sensorimotor Road to Artificial Intelligence.' He cited Moravec's paradox from the 1980s: “It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”

Sure, AI models like GPT-4 can pass the bar exam - there is copious amounts of training data on the web to set these AI models up for success. But an AI-powered robot cannot slice an onion or scramble eggs, something a 12-year-old can do. And that’s despite being trained in pristine, spacious laboratory environments.

In the real world, you must factor in varied home layouts and personal preferences. You also have to factor in the randomness and occasional chaos in our environments, like a toddler zooming around on a toy car or blowing bubbles in the kitchen as you try to prepare dinner.

The training data needed to accomplish these sensorimotor tasks, much of which comes from the formative sensory experiences in a child's first few years of life, is precious, personal, and mostly inaccessible.

Researchers at NYU are now capturing that data - the training data equivalent of pure gold - to train AI to learn language more efficiently, like our little ones.

Perhaps this data could also be used, as Professor Malik suggests, to help robots dice tomatoes.

One Trick Pony

We know scientists are trying to improve robots’ sensorimotor skills to be more effective at physical household tasks. But are there any robots that can take even one of these demanding chores off our plates? We’re desperate here!

Well, that’s precisely it — scientists and technologists have been able to introduce robots that can semi-successfully do a single task - like make ice cream or cook ramen. However, the robots’ often flawed execution of even a single task, along with their hefty price tags, make these single-task robots unviable for the broader market.

Let’s look at an example.

In 2019, FoldiMate debuted a laundry folding machine! However, there were a few drawbacks:

1. You had to feed each item individually into the machine to be folded.

2. The machine couldn’t fold sheets, towels, or baby clothes.

3. The machine cost $980.

FoldiMate ceased to exist in 2020.

Multi-Trick Pony

Don’t despair. A team of Stanford scientists recognized that a good household robot would be one that can perceive variations in its environment and learn different tasks according to the needs and particulars of the household it operates in.

This team is developing Mobile ALOHA, an AI-powered robot that learns and imitates human movements. The Stanford researchers strap themselves onto the robot and can teleoperate it through a given task. After doing a particular task about 50 times, the robot can imitate it autonomously.

Through this training framework, Mobile ALOHA has learned how to put away a cooking pot in a cabinet, call an elevator, push in chairs, sauté shrimp, clean up a beverage spill, and give high-fives.

Downsides? Yep.

The prototype of Mobile ALOHA costs $32,000. It’s also clunky and could be a tripping, choking, electrocution, and fire hazard. I’m sorry to say it is neither demure nor cutesy.

But it’s early days! Future versions of the robot may be smaller, exhibit greater freedom of movement, and be more manageable for non-experts to operate.

What’s to Come

While we wait for Mobile ALOHA to iterate into something more toddler-in-kitchen-friendly, other types of household robots might start rolling into homes.

Over the last few years, Samsung has showcased household robot Ballie at major trade shows. This spherical yellow robot acts as a smart home assistant on wheels. Like our existing smart home assistants, Google Home and Alexa, Ballie can control connected devices in your home.

Ballie has a camera and can send pictures and updates about pets and family members when you are not home. It also features a projector that can project your doorbell camera, the inside of your Samsung oven, or recipe videos as you cook.

But let’s come back to the question of the hour: can Ballie fold your laundry? Nope. It can probably project your favorite Netflix show to watch as you fold your laundry, though.

Breaking news: Neo

Over the weekend, the topic of household robots went viral again, with the company 1X previewing their humanoid robot built for the home, Neo.

I hate to judge a bot by its cover, but that thing is creepy.

After perusing the 1X website and social media accounts, I found little evidence of what this robot can do - or what it is being trained to do - besides what is shown in the video: have silent conversations, hand you your backpack, and embrace you in a cold, hard half-hug.

Neo is still in development, so here’s hoping that the ready-for-market version looks less like Slenderman in a tracksuit and can fold a T-shirt!

The verdict

As I researched this topic, I was hopeful I could tell the ammi.ai family that our salvation had arrived in AI-powered laundry-folding robots.

But alas, we aren’t quite there yet. As we covered, the mechanics of folding laundry are very complex and will require extensive training and engineering feats for robots to achieve.

Second, Ammis are particular about their homes. We like our kitchen and cooking handled a certain way, our laundry folded and put away a certain way, and our homes cleaned a certain way. A big part of what makes a house a home is something a robot will never have: a heart.

The good news is there are lovely people out there who are exceptionally talented at household tasks, can put their hearts into it, and would appreciate the work.

So, if you’re willing to invest hundreds, maybe thousands, in a robot that could fold your laundry, maybe opt for a person with a heart who will do it better than any robot could.

Fruitful Harvest: fresh AI fruits from the community garden.

Here are a couple of fresh picks for you!

🍐 A Nostalgic Throwback or an Unsettling Warning?

The 1999 Disney movie Smart House is about a family that wins a futuristic home equipped with an AI household assistant named PAT. Things go awry when the AI tries to take on the role of Ammi.

🍇 Resources to Manage Family Screentime

Many parents try to curb their family’s screen time as much as possible, but it’s hard to draw boundaries and it’s even harder to enforce them. Here are some useful Family Tech Planners by age group, developed by Common Sense Media, to help guide you in creating screen time rules at home.

Planners by age: Ages 2-8, Age 9-12, Age 13+

🍉 AI is not that good at math

As students return to school, they may use ChatGPT or similar AI tools to help with homework. A word of caution: be wary of AI assistance with math homework.

Cultivating Conversation: your insights nourish the ammi.ai garden.

We 💚Your Feedback

We 💚Your Insights

Thanks for spending a few of your precious, precious minutes with us.

See y’all soon!
Ruqaiya
Ammi by day, Ammi by night