In your job, when will common knowledge be wrong?

American surgeons in the late 1800s prided themselves on wearing blood-stained coats. It was the symbol of a busy practice. Atul Gawande wrote about this tradition in an article for The New Yorker:

Surgeons soaked their instruments in carbolic acid, but they continued to operate in black frock coats stiffened with the blood and viscera of previous operations...Instead of using fresh gauze as sponges, they reused sea sponges without sterilizing them.

After researchers proved the life-saving benefits of antiseptics, it took almost a century for most doctors to start scrubbing their hands and sterilizing their instruments. Today, of course, a surgeon prides herself on wearing a sparkling clean, white coat.

The point isn't to pick on healthcare professionals. Over the course of two centuries, the most widespread beliefs in any profession are likely to change — even flip entirely. To quote the great computer scientist Marvin Minsky:

All knowledge has a half-life: the time it takes for what you know to become wrong or redundant.

As a thought experiment, imagine you're walking in the shoes of someone who's doing your job 25, 50 or 100 years from now. Speaking or writing from this person's perspective, describe a belief that's widely accepted in 2017, but obviously wrong in the future. Why did this belief change and what idea took its place?


Here's what I think a software engineer in 2042 might say:

In 2017, consumers could speak to devices with human-like personalities, but few of the personal assistants had persistent emotional states. In fact, it was funny to think about a device expressing emotion. In the following decades, software engineers recognized the utility of feelings. Today many consumer devices have emotional profiles that persist from one interaction to the next.

Here are three practical reasons we incorporate emotion into consumer AI:

  1. Understanding: Emotional affect makes it easier for us to derive meaning from language. For example, imagine that someone approaches you and makes the following statements with an identical tone and emotional affect. It would probably take you longer to parse and react to the messages:
    • "The house is on fire."
    • "I love you."
    • "Would you like a piece of cake?"
  2. Purpose & Meaning: Our interactions with other humans have purpose because we know that the things we say and the choices we make have lasting emotional impacts. Giving machines an emotional state — albeit a very simplified, virtual one — makes us value our interactions with them.
  3. Awareness of Environment: Emotions such as fear and anxiety help machines to navigate their environment safely. For example, the ability to experience pain related to pressure and heat helps a robot in a factory avoid dangerous obstacles. This discovery dates back to 2016, when German engineers developed an artificial nervous system.