Recently I initiated an online chat with Apple support because I had a problem with my iPhone. The agent, Jayson, told me to turn the device off and on, but this didn’t fix the issue. Jayson then suggested I have the phone serviced by Apple, and arranged an appointment with the ‘technicians and geniuses’ in a local store. I thanked him and this exchange ensued:
You’re most welcome! Thank you for your kindness, understanding & patience. Much appreciated John! Was I able to help you today?
By the way ~ May you have a prosperous 2019!
Thanks for your help. Bye now.
I really appreciate you John, Don’t forget to eat your breakfast, lunch & dinner okay,? Its been my pleasure assisting you today, again my name is Jayson. Have a great day and take care always~
All is well! You deserve the best in life and Cheers for a great 2019!
Running a very elementary Turing test on this dialogue reveals that Jayson is an artificial rather than human intelligence. He (it) says things that no adult human agent would say in this situation. Yet Jayson’s mistakes are almost human, like those of a toddler learning to use language. He writes his own name phonically rather than conventionally. He can’t deal with my question – beyond the standard remedy of turning the device off and on – and so passes me on to the adult Apple technicians and geniuses. Unlike a traditional machine, he expresses affection and care: he wishes me a prosperous 2019 (in March) and reminds me to eat regularly, as his parents might have told him. No human programmer would make these mistakes; Jayson’s algorithm needs to learn to adjust its warm, supportive language according to the season and the relationship.
Within a few years, it will become much more difficult to know whether one is ‘chatting’ to a human or machine agent (although we shall probably assume the latter). And this of course raises the question of the nature of knowledge: given appropriate technology, could a machine learn so adequately that the distinction between human learning – a activity of the embodied mind – and machine learning is elided? In Bladerunner 2049, there is no evident difference between the humans and the replicants. In Cultural Literacy (1988), E D Hirsch states that researchers in artificial intelligence have concluded that knowledge is the key component of all cognitive skills: ‘Once the relevant knowledge has been acquired, the skill follows’. Machine learning, according to this view, is not a mere simulacrum of human learning, but its paradigm.
This view clearly has very profound implications for educational policy, and aligns with current influential views on the teaching of language.
To be continued …