AI and imagination

Does society need machines that show intelligence, empathy or act like human beings in other ways? I think that we are best served by well designed computers and good interaction design that makes clear the distinction between the person and the machine.

Much of the enthusiasm for AI-based human-computer interaction assumes that people desire a seamless blending between the human and the machine. Demonstrations of intelligent human-machine interaction, now or in prospect, are not effective enough to persuade me that we really need AI interfaces.

Understanding and accountability

We human beings understand each other better than we understand machines, systems, and technologies. We are well adapted to human-to-human interaction. Part of that interaction involves ambiguity, covert signalling, and a host of other soft-edged, contingent and multilevel communications protocols. In our conversations we pose questions, utter prompts, proffer answers, and probe further depending on circumstances. Machines designed to act in this way invariably reach limits and fail “the Turing test.”

Humans are not well served by an interface mimicking human intelligence anyway. Dealing with a person directly provides a conduit to whoever is accountable. As anyone knows who’s been frustrated by an automated checkout, ATM, ticket machine or voice-activated phone service, machines that perform as human surrogates cloud that accountability. Who do you turn to when things go wrong?

Sub-intelligent interaction

Humans are also adept at understanding tools, devices and technologies. We don’t need our machines to offer fake human attributes in order to interact with them. According to Apple’s current website for iOS 9, “you’ll get intelligent suggestions for what you might want to do next — even before you ask.” That’s fine for sub-intelligent interaction, but think of the human effort required to check, cross check and ultimately trust interventions of a more serious kind: steering your car, transferring investments, or advising on your career.

Not all AI attempts human-like responses from machines. So-called knowledge-based systems (KBS) focus more on invisible computer functionality demonstrating human-like expertise in solving particular problems or performing expert tasks (medical diagnosis for example). There’s no suggestion here that KBSs need to exhibit personality traits.

But it’s the sympathetic, conversational interface (think of the movie Her) and the image of the humanoid robot (Ex Machina) that capture people’s attention when they think of AI.

Imaginative interaction

Supposedly intelligent interaction renders literal and prosaic the human capacity to imagine. I can pretend that my car has personality or that Siri is sentient, but not all the time.

Putative AI interactions come across to me as an attempt to spoil our imaginative play by persuading us that what’s at the interface is really human. Artificial intelligence sullies the important play between imagination and reality.

References

• Coyne, Richard, Michael Rosenman, Anthony Radford, M Balachandran, and John Gero. 1990. Knowledge-Based Design Systems. Reading, Mass.: Addison-Wesley
• Kurzweil, R. (2005). The Singularity is Near: When Humans Transcend Biology. New York, Penguin.
• Turing, A. M. (1995). Computing machinery and intelligence. Computers and Thought. E. A. Feigenbaum and J. Feldman. Cambridge, Mass., MIT Press: 11-35.
• Weiser, M. and J. Seely Brown (1996). “The coming age of calm technology.” Xerox Parc Report http://www.ubiq.com/hypertext/weiser/acmfuture2endnote.htm.

Notes

• “To erase the line between man and machine is to obscure the line between men and gods.” Ex Machina movie trailer
• Uber automated online car hire/drive service
• A short interactive movie about a lost iPhone by Zuo Migni (MSc Design and Digital Media 2014-15)