- Details
- Participants
- Notes
- Uncanny Valley
- Data
- Trade-offs
- Henn-na Hotel in Japan
- Study
- Project Comparison
- Resources
- Categories
Details
Podcast: 
Episode: AI Failure Series: Falling into the Uncanny Valley
Participants
Ron Schmelzer
Ron Schmelzer
Kathleen Walch
Kathleen Walch
Notes
Uncanny Valley
- It is the thinking about the relationship between something that resembles a human and a human’s emotional response to it.
- The uncanny valley is when the human is made uncomfortable.
- Examples
- Industrial robot: doesn’t resemble a human, not creepy at all.
- Wall-E: Sort of human, but it’s cute. Not uncomfortable.
- Animatronics: Starting to resemble a human. A little uncomfortable.
- Sofia: It looks human, but there’s something off, and it feels uncomfortable.
- Fully human response and look - humans are out of the uncanny valley and comfortable again.
- Humans become resistant to interacting with these robots in the uncanny valley.
- When people fall into the uncanny valley is different for everyone.
- Having been in uncanny valley results in degraded trust, and it is hard to get out of the valley once you are in it.
Data
- It doesn’t need to have a physical manifestation.
- Uncanny valley can happen with data as well.
- Examples
- Customers who purchased this also purchased. Most people are comfortable with this.
- You may be interested in it. A little weird, but I still will interact with the service.
- Do these interest you? And it’s something they had only thought of once, and they wonder how it did it know I was interested in that. At that point, you are in the uncanny valley.
- It generally comes from other data like location tracking, calendars, etc.
Trade-offs
- Convenience vs. Privacy. The ability to provide more convenient features largely depends on the amount of data the system has about the user. But then, having all that data is a privacy issue.
- Transparent vs. secure. The ability to be transparent and open about how you use data and the underlying systems is counterbalanced by the need to keep some things hidden for security’s sake.
- Value vs. privacy. The ability to provide more valuable, unique services is also based on the type of data you have about the user. Having that data, like PII, is a privacy issue.
- Convenience, value vs. trust - The above benefits vs. the fact that users don’t trust the people or organizations holding the data.
Henn-na Hotel in Japan
- The idea was to cut labor costs by mainly having robot workers in the hotel.
- Roles
- Check-In
- Humanoid - spoke many languages.
- Velociraptor
- Concierge
- Baggage Handler
- Wait Staff
- Rooms door opened with facial recognition.
- An in-room robotic assistant to help with lights and room service. They called it Churri.
- Lasted for two years
- The robots crossed into uncanny valley.
- Reasons
- Churri was chatty and woke people up in the middle of the night.
- The concierge service couldn’t handle a lot of the requests.
- Baggage handlers couldn’t open some guests' rooms.
- Check-in robots couldn’t make copies and do other simple tasks humans do.
Study
- One of their studies shows that 61% of Americans don’t trust robots.
Project Comparison
- Airline self-serve kiosks
- Humans use a touch screen to check in and print boarding and luggage tags.
- People prefer this over the old way of standing in line and waiting for a representative to do it.
- The human is in control.
- If the airlines went to a humanoid robot that moved instead of kiosks, then:
- Most likely a failure due to falling into uncanny valley.
- Most humans prefer returning to standing in line over dealing with a mobile humanoid robot.
- The human is no longer in control.
Resources
Cognitive Project Management for Artificial Intelligence
