sad robot needs a break
A robot comedian? Might not be as far-fetched as you think PHOTO: JD Hancock

Will robots become so strong and smart, so agile and creative, and so ambitious that they will ultimately rule the world?  

Don’t be ridiculous. 

Or so I blithely thought until I researched this article. Here’s what I found that is a bit unsettling for humans.

Myth 1: Robots Are Strong, but Not Agile

Two years ago, the researchers at Stanford’s Biomimetics and Dextrous Manipulation Lab designed a tiny geckobot that pulls up to 2,000 times its weight — the equivalent of a human being pulling a whale across land.

Robots are clearly stronger than humans. They are not however, as agile as humans.

Wrong.

The École Polytechnique Fédérale in Lausanne, Switzerland has invented a robotic gripper that can not only carry 80 times its own weight but is also dexterous enough to pick up an egg or a single piece of paper. This gripper is the first to incorporate electroadhesion and sensors that allow the robot to adapt itself, representing a new approach to agile robotic manipulation.

And robots are increasingly performing all kinds of fine motor skills allowing them to perform jobs once done by humans — and to do so without pause. Like the sorting and assembling robots in this video, oddly reminiscent of Ethel and Lucy in the Chocolate Factory, that never seem to weary or get bored (at least as far as humans can tell).

Myth 2: Robots Are Smart, but Cannot Determine Intent

Robots are smart enough to replace many of the jobs human knowledge workers are doing today. And Masayoshi Son, CEO of mobile company Softbank, predicts smart robots will outnumber the human population by 2040. 

Artificial Intelligence (AI) bots are already able to transcribe speech better than professional transcribers, spot cancer in tissue slides better than human epidemiologists, or do due diligence on M&A deals better than legal techs. And AI is creating bots that can even teach themselves to be smarter. DeepMind researchers have already seen their AI product achieve credible results teaching themselves to play video games.

Robots are super smart and getting smarter, I get it. But they can’t really interpret intent.

Wrong again.

Consider the AI-infused Harry Potter sorting hat: you place it on your head, tell it a few things about yourself and then it sorts you appropriately. (Of course, the “real” sorting hat is not just smart, it's an enchanted hat that sorts students into Hogwarts houses by discovering their intent.) 

The AI version runs on a Natural Language Classifier to interpret the intent behind how people describe themselves. Deciding what words associate with a specific house "sets a ground truth," like the characteristic honesty for the Hufflepuff house. The hat adds more characteristics on its own using Deep Learning, “a branch of AI when machines learn to complete tasks on their own,” to associate keywords and expand the number of ground truths. 

The hat gets smarter over time by listening and observing — scanning the web — a behavior that mimics how human muggles learn these days.

Myth 3: Robots Can Learn, but Not Self-Evolve

We might reassure ourselves by thinking robots are strong and smart because we made them strong and smart, and can learn because we programmed them to learn. They can’t actually make better next generation versions of themselves.  

You know what comes next.  

A Norwegian robot has learned to self-evolve and 3-D print the next-gen of itself. Experts at the University of Oslo are using Generative Design “where artificial intelligence programs — not humans — innovate new products.” 

This video provides an eerie, and perhaps amusing, look at how it’s possible for robots to design and evolve. Robot “Number Four” is made of sausage-like plastic parts linked together with servo motors. It tries out different gaits, attempting to figure out the best way to move from one end of the floor to the other. 

By constantly monitoring its own progress and comparing it with previous attempts, it visibly improves over time. By simulating a thousand robot generations, the computer can come up with a working model within a couple of hours and instruct a 3-D printer to make a real world version of this “pinnacle of evolution.”

And with the rise of 4-D printing — printers that both print and assemble — there’s no reason why robots won’t be able to design, print and assemble better versions of themselves in the future.

Myth 4: Robots Can Be Creative, but Not Emotional

Strong, smart and self-learning — but can robots also be creative? Yes. In an impressive compilation of 59 things artificial intelligence can do today, robots are capable of painting a pretty good van Gogh, writing and publishing poetry and composing music.

But the secret sauce to humans is not just our creativity but our delightfully imperfect emotions. Robots can’t do emotions, right? 

As the article "Welcome to the Emotion Economy" tells us, technology is quickly gaining the ability to understand — and anticipate — human emotions. This is a seismic advance since emotions are at the heart of the human experience, “playing crucial roles in communication, social bonding, even our decision making.” 

And so, we are seeing the beginning of the "emotion economy," a range of interconnected software and services which seek to fulfill the promise of emotionally aware machines. 

While a relatively new branch of computer science, affective computing stems from work initially done at MIT Media Lab, which moved into commercial applications that require a level of EQ (emotional intelligence), such as companion robots for seniors and PTSD treatment in soldiers. 

Don’t Make Me Laugh

It turns out robots are stronger, smarter and more emotionally mature than I thought. Fear not, though, robots cannot breach the final frontier of humanity. Robots cannot be funny. 

Unless I'm wrong again ....

Joke-Telling Robots put it perfectly: “Humor requires self-awareness, spontaneity, linguistic sophistication, and empathy. Not easy for a robot.” Yet AI researchers are working to create robots that detect various shades of wit from humans and respond with their own quips.

Serious work on humor for robots is being conducted at universities including Purdue Polytechnique, the Association for the Advancement of Artificial Intelligence held the first-ever symposium on artificial intelligence and humor a few years ago, and comedic robots even made a splash at SXSW Interactive. 

At this point, robot humor remains predominantly template-based and dependent on learning from familiar jokes. 

But it is moving inexorably towards the ability to generate humorous responses — jokes — to situational humor. And when that day comes ... the joke's on us.