Interesting Collaborations Between Robots and People
Posted in: UncategorizedI’m a long time Industrial Designer, the current Graduate Director of the MID Graduate program in RISD’s Industrial Design department, and a Faculty in Residence at the Autodesk Build Space in Boston.
As a resident, I proposed to expand on Autodesk’s work of creating collaborative cloud-based software environments for creating, product part drawings for design and manufacturing by investigating the stages just before project teams (designers, engineers, etc.) decide to commit to mass production.
Duets with Machines
A few months into my residency I was invited by Moving Brand‘s CCO and co-founder, Jim Bull and Interaction Designer Dave Cameron to join a “Duets with Machines” panel discussion to talk about my Build Space experience and to consider a future of working with robots.
The Duets with Machines panelists (left to right) Jim Bull, Ivy Hu, Caroline Sinders, Andy Law and Dave Cameron. Photograph © Moving Brands Dave: “[Jim and I] are interested to understand and investigate what does it mean to collaborate with machines to generate creativity? As it is becoming easier to collaborate with machines, from conversations to automated processes, we want to explore what this means for our own practice and how is it influencing other’s creative work.”
This article is an expanded and edited version of that presentation and the unexpected discussion that followed. The whole conversation seemed to not just come down to how people perceive ‘robots’, but more importantly, the culture that created these machines.
Areas of Opportunity
-Collaboration between experts, such as scientists and civic innovators, is very difficult because they have very little specialist-shared vocabulary.
-Communication between medium sized factories and people batch producing projects is challenging because it requires a lot of time, which rapidly increases everyone’s costs.
-Combining CNC production services (such as waterjet and laser cutting) and garage shop methods (welding and assembly) into viable product manufacturing is underused because of the broad skill set needed.
My approach to investigating and trying to solve these problems is by creating specific tools for helping (in a machine duet) people transition from ideas and prototypes into batch runs of products.
The 300 Batch Narrative
For most people, making a few prototypes is easy, and for others (myself included) producing a few thousand products or components is also easy, while making only 300 hundred is strangely quite difficult. These struggles take the form of various stumbling blocks: making and setting up new tooling is expensive, learning or teaching a new process takes time, getting high quality can require lots of corrections, selling and testing takes just as long as 3000, the final price is always high.
But the benefits of making just 300 items go beyond exploring if people are interested in your idea (the MVP concept) or proving that it can be manufactured. Growing slowly can be a valuable experience because you can test to see if your product idea will scale in the way that you imagined. We will see that it rarely goes according to plan.
Supporting people wanting to scale through the 300 barrier is a perfect opportunity for people and robots to collaborate, as these people will use tools that help them work quicker, more accurately and longer.
Let’s start by looking at some examples of new tools that are already being developed in an area that I have been calling Augmented Hand Tools. But, in this first example, it’s more about legs.
Augmented Legs
A simple illustration of what we might all see as a duet comes from Superpedestrian, a tech startup that is working on the Copenhagen Wheel, a replacement for your existing bicycle’s rear wheel that includes a motor, battery, and a ‘robot.’
Andy Law riding a bicycle fitted with Superpedestrian’s Copenhagen Wheel. Photograph © Christina Galvez The Copenhagen Wheel doesn’t require any extra controls because it works intuitively, anticipating when the rider wants more energy and adding it to normal peddling as the software within the robot measures when the rider will want more speed or power. When I was able to test the Wheel, I felt like I had super powers. It was fascinating how quickly I forgot that I was involved in a duet and had a robot helping me out. It took me at least 20 minutes to stop zooming about and showing off.
Horatio Han, a designer at Superpedestrian, told me that people seemed to be buying the Copenhagen Wheel to allow them to bike 20 minutes to work. Without the wheel, this commute would take 35 minutes.
The Flow of Work
Ilan Moyer and his colleagues at Shapertools have created a handheld robot wood router that corrects your machining mistakes. With very little practice, I was able to cut a perfect circle out of plywood sheet. A user could quickly make all the parts for the perfect table or shelves, cut to perfection.
The way that you work with this tool is novel because it doesn’t use the traditional process of wood routing. The user doesn’t need to draw something with a computer, nor do they use a machining guide. The robot has a small screen that shows the user a second view of the material so that you can place and size onto it predefined shapes of what you want to create.
Kelly Lohr, a RISD MID student using a Shaper Origin prototype. Photograph © Andy Law and Shapertools The design of the workflow has deliberately changed. You don’t have to be linear, and you can work things out or adapt them as you go and still make perfect components. Once you are happy with your prototype, you have essentially created your templates that can be used again.
Pencil Duet
My original goal was to create a machine duet with a person using a pencil that would help them reveal onto paper, a photograph taken by their phone. But it became my attempt to create an Augmented Hand Tool.
I have essentially created a pencil that knows where it is on a piece of paper and simply alters how much contact you have with the paper. This has the effect of making heavier or lighter lines, and you can do what you like with that. You can lean in and apply more pressure or go over it twice or even ignore part of an image. I can also switch to reacting to a specific color (just red, green or blue), increase the overall contrast, or even invert the entire image.
One of the early augmented pencil drawings. Image © Andy Law The most interesting thing about this is the relationship that I’m developing with the pencil. We work together, but we also seem to disagree and ‘argue’ when the pencil goes wrong. It glitches a lot, but I’ve grown to like the imperfections.
I’m still working out what this particular collaboration means and how to best use it exactly. However, it seems like my robot pencil and I will create some unusual drawings on the way that will be very different from my usual sketches.
Bonded Conversations
Before the panel discussion, I met and talked with Ivy Hu, the lead Experience Designer at ASAPP, a startup that is applying Machine Learning and Natural Language Processing to selected call centers.
She described how the customer service chatbots that they are developing were an extremely effective experience when the conversations were kept to very specific subjects that had a limited number of outcomes. Another deliberate bounded simplification was the use of text messaging for the conversations, as this was low in errors and avoided the many problems of speech recognition. I’ve had similar successful bounded conversational experiences with what I believed to be real people when I was buying a mattress, trouble shooting my vacuum cleaner, and upgrading my home’s internet connection.
Ivy went on to explain that the chatbot’s main purpose is to connect the user with a real person who was an expert in the right problem area and that, by designing the experience around a familiar non-expert receptionist’s suggestion of a potential specialist, this cleverly avoided the requirement to understand and answer questions with high confidence and again avoided a bad experience.
Designing Duets
These four duet examples help to create some simple design guidelines as they:
-Highlight that robots are brilliant specialists that perform well in carefully bounded situations, like playing Go or freeway driving in California.
-Show that people are great generalists who can relate, adapt, and apply a wide variety of influences to any situation.
-Indicate that people are highly adaptable and can revaluate and critique results and outcomes against complex sets of constantly evolving values.
-Support the idea that robots and people work better together, and that the output of these duets can create a better experience than a robot or a person working individually.
Augmented Hand Tools sit within this duet area and could be defined as tools that extend your existing skills into super skills by introducing specialist assistance, such as auto correct/suggest/real-time analysis, all the while allowing you to experiment, play, and learn. These duets can train you, help you advance your skill sets, sharpen accuracy, increase stamina by making things easier or more difficult, and even record your actions so you can replay them or share them later.
Augmented Hand Tool Benefits
I was lucky enough to meet Pratap Bose, Head of Design for Tata Motors. He told us that even though they recently reduced the car development time for their TaMo Racemo from 72 to 22 months, they are interested in evolving tools and processes to help the company innovate and get cars onto the market even more quickly. This drive began after they created the $2000 car Tata Nano. Even though they spent six years developing this car, it didn’t sell as well as they had hoped for a variety of reasons.
One big reason was that the project team’s understanding of modern India was incomplete, but they only realized this after the car didn’t sell as well as they had projected.
Having new tools would help Tata’s designers innovate even faster, allow them to use production processes that are efficient in lower numbers, reduce the cost of development, speed the time to market and importantly, allow more prototyping.
Increasing the frequency of quick testing with iterations of prototype cars will involve more end users, which will dramatically reduce the possibility of Tata making a mistake that could result in people not liking their products. Tools for scaling up through 300 and not jumping from 3 to 3000 will make sure this problem won’t happen again.
My Perspective
There is one other reason why I would like to see tools and processes that allow more people to get involved in the making and scaling up of products. People disliking your product is one problem, but people using your product in unexpected ways can be another.
There is an inescapable and logical connection between Industry and Society because products reflect the society or culture that created and that use the product. And these two are not always (almost never) in agreement.
Reflecting a culture can mean empowering it, but it also means amplifying and exposing it. I have some quick examples to illustrate this.
Industrialization
Industrial Design and Engineering have a long history of working with machines.
Much of this work originally focused on controlling the product and forcing the workflow to maintain a standard and quality of production that, until automated machinery entered the manufacturing process, hadn’t existed.
This approach was intended to ensure that products were produced efficiently, affordably, and could be maintained easily out in the ‘real’ world. Notably, the process reflected the linear and top-down governance model of industrial ownership, which was largely a reflection of how society was structured at the time.
Flexible Society
Starting in the ’50s, there was a global move towards a less rigid, more systematized and flexible approach in American society that became reflected in how things were made. We now know this form of production as Lean Manufacturing, and it has continued to evolve to try and keep up with the increasingly diverse aspirations and makeup of today’s society.
Some people see the future of increasingly flexible manufacturing as a big ‘black box’ making machine, where small teams of engineers control multiple robots that perform most of the production work.
Controlled Environments Meet the Real World
Undoubtedly, this flexible approach to manufacturing works well inside a controlled environment like a factory. But more importantly, is this an accurate reflection of our current society?
Another panelist at the Duet with Machines talk was Open Labs researcher Caroline Sinders, who had some extreme but interesting digital space examples from her investigations of (what I am describing as) online products leaping from prototypes to products. Here are two examples:
Microsoft recently made their experimental machine learning chatbot, Tay, available online so that it could learn conversational skills from users in real time. Within 24 hours, Tay demonstrated its capabilities by making racist comments. It was promptly taken offline.
Google’s image search has similarly been ‘caught’ producing surprisingly racially biased results that have proven to be difficult for Google to explain or correct.
These examples prove how some unexpected aspects of the cultures that created them, and how people are subsequently using these tools, products, and research projects is clearly being reflected.
What is the future for Machine Duets?
I believe that we will see more duets such as Augmented Hand Tools that will support scaling up through batches of 300. Society is becoming more connected, and we are more aware of our diversity; this is not only the most probable future but the most important. We desperately need to get the means of production into more diverse hands.
So, I’d like to finish with another question for everyone creating tools:
