Let’s face it: Who wants to do boring and repetitive work?
Probably not you and me or most other people if you ask them. Yet, many of us are faced with repetitive tasks daily like writing the same kind of emails or reporting in to the boss with information he or she may already have access to.
Instead of insisting on doing those tasks, perhaps it’s time to start thinking about how to automate them?
I asked Mads Rydahl, the former senior director of product design at Siri (yes, that same assistant that makes you talk like a robot to your iPhone), and he expects to see an increasing amount of self automation in the future.
Basically it’s about teaching a computer to do your most repetitive work tasks — something which artificial intelligence is already quite capable of in many cases.
“With self automation you’ve probably gone one step towards making yourself redundant. But you have also become more effective — and perhaps you could do the work of your colleague instead of being fired,” he says.
The reason he talks like that is that he’s seen first hand how artifical intelligence better known as machine learning is becoming increasingly capable of doing very specific tasks like understanding voice commands or to find patterns in large data sets.
Since Apple bought Siri in 2010 and put it inside the iPhone, Rydahl has continued his work to build artificially intelligent systems that are able to process human language.
Today, he is spearheading the semantic search engine Unsilo, which helps scientific publishers like Springer Nature and scientists to discover similarities across millions of scientific papers and drive new discoveries.
Solving the 21st Century’s Science Overload
As a research scientist it’s becoming increasingly complex to “stand on the shoulders of giants” and find the right research to build upon with future investigations. Because of what is known as the ‘21st century science overload’ it’s hard to keep up — each year scientists publish about 2.5 million new research papers and that increases by over three percent each year.
If you’re doing research about diabetes in overweight girls, the Unsilo search engine understands the meaning of the concept and finds related research papers with other variations of that particular concept — a feat that is getting very hard to do manually.
In other words it’s possible to make sense of data sets so huge that they’ve previously been off limit for humans. Or perhaps they’ve required a lot of manual work to untangle.
In the future, researchers may have to be more creative with the kind of tasks they assign to research assistants. The most rudimentary ones are about to be automated if it is worth the investment.
Another example is the thousands of documents relating to climate change that are published every year. Today a lot of manual work is put into categorizing these documents (is it about climate change or the weather next week?) but that may soon enough change. Unsilo is already testing an AI system that’s supposed to automate a large part of this categorization in collaboration with an unnamed large organization behind the UN Sustainability Goals.
What does that mean? Well, for one it means that fewer people will be needed in the future to do those repetitive categorization tasks. At first it might not seem so, but that’s only a matter of time according to Rydahl.
Once the software starts running, each worker will train it to take over his or her work tasks little by little.
“They take the knowledge from their heads and put it into the tool, we’re building. When the software knows how to decide for itself, then they’ve formalised that knowledge,” Rydahl says.
Look For Deja Vu
Self automation is something which Dennis Mortensen can talk about as well. He’s the CEO of X.ai — an artificially intelligent assistant that books meetings for you through email almost like a human being.
His self automation ‘aha moment’ came after selling his previous company in 2013 when he counted the amount of meetings he had held the year before.
“I had held 1019 meetings in a year and rescheduled 672 of them and I had done all that by myself,” Mortensen says on Skype from New York, where his company resides.
Instead of following the path of most other executives and hiring a human assistant to schedule meetings in the future, he opted for creating an artificially intelligent bot to do it instead.
The trigger was simply a hunch he got from looking at his own data.
“I had the feeling of repeating myself and in doing so I immediately got the sense that there were some patterns which perhaps you could learn by using machine learning,” he says.
The story which led Mortensen to create the bot Amy is a fascinating one, which I’ll save for a later post.
Now it suffices to say that “she’s” booked millions of meetings for people by understanding and responding in human language in emails — like a human secretary would. Except she doesn’t take breaks, sleep or ask for a raise.
The message I want to convey with these examples is simply a suggestion to look up from the keyboard or whatever tool you’re using for your work.
Ask yourself: How often do you feel that you’re repeating yourself? Do you think that many other people in your position also experience that?
If it it’s too often, do something. Find a way to automate the task and get a promotion. Or look for something more valuable to do before someone else automates the task.
Personally, I’ve already automated the task of bookkeeping in my company. Whenever I get a bill, I just send it as an email to Roger, “who” automatically scans it and schedules it for payment on the proper date. At the same time I forward the email to my accounting system Dinero, which then uses AI to categorize it properly.
All it takes is about 15 seconds of my time and no need for a bookkeeper.
What do you think about the ideas presented in this post? Do you see a potential for self automation in your work and if so, how?
I appreciate all comments and feedback. If you enjoy this post and want to follow the blog series, you’re welcome to subscribe below.