I realized Artificial Intelligence (AI) had left the world of science fiction when scheduling a meeting for my boss, Dr. Darrell West. I am Darrell’s assistant in his capacity as Vice President of Governance Studies. I assist Darrell with daily office logistics like scheduling meetings, greeting guests, and arranging travel. In addition I provide research support for blogs, papers, and books. My position requires me to balance priorities, intuit preferences, and use interpersonal skills. Coordinating Darrell’s schedule is a daily function of my job and I recently learned that in this capacity, I could soon be obsolete.
Meeting my potential replacement
My introduction to Amy was a pro forma email requesting a meeting with my boss from a technology policy expert. Amy was copied on the email to help schedule the meeting. At this point I had no reason to believe that she was not human. Her first email to me seemed normal. But I realized something was off when Amy sent follow up messages to me late on a Friday and Sunday night. I followed a link in her email signature to the website of a company that designs artificial intelligence to schedule meetings. Amy was not a human personal assistant and without the link in her signature, I would have never guessed that she was AI. Amy was polite, persistent, and an effective scheduler. Amy is likely better than me at some tasks.
The future of the AI assistant
Amy demonstrates a new chapter in the interaction between humans and AIs. In this case a computer program not only replaced a human but also did so surreptitiously. This is a departure from the classic story of a computer replacing a human. When the ATM replaced the bank teller the change was clear. AI has achieved a level where it can mimic appropriate email language and etiquette quite impressively. The value of this innovation stems from not knowing that Amy is a robot. Presently, AI assistants communicate primarily with humans. If a human is aware they are scheduling with a robot, this might influence the correspondence negatively. The human could take offense that the message writer did not think the correspondence was important enough to be handled personally. The human might even be slow to respond or outright ignore the AI’s emails with the reasoning that you can’t hurt a robot’s feelings.
Amy gave herself away by emailing me on a Sunday night. It takes a human to know that you can’t expect an email response late on the weekend. How can you build a computer program to know that? Of course, it is easy to check a box in Amy’s preferences not to send emails on Saturday or Sunday. But some preferences are more complicated to program. For example Darrell is a huge Patriots fan. It takes a human to know that I should not bother him with minutia on the Monday after a bad loss.
AI assistants have the potential to drastically change this field. However, I have confidence that my job is safe. Although AI technology has improved by leaps and bounds, they aren’t a major threat to my position because it requires critical thinking and analysis. Robots are great for routine tasks but it will take time before AIs properly assess complex workplace decisions.