If an artificial intelligence is meant to be thinking and feeling, then it makes sense to understand it as a social being, one which learns in a community of learners, teachers, and others.
This might seem obvious. Just as a human being can become human insofar as they are embedded in a community and human language, so it should hold for an artificial intelligence. The current approach to AI is fixated on a solitary learner who is fed data and then is meant to approximate some degree of intelligence. This is mainly due to the disappointment with the project of a General AI or Artificial General Intelligence and the retreat into Narrow AI research.
Still, what we are dealing with is usually unsophisticated number crunching which attempts to solve problems through brute forcing large amounts of data, and thereby finding efficient patterns which are not obvious when dealing with predetermined rules or partial data.
Reggio Emilia principles vs. Intelligent Agents
Reggio Emilia, a town in Italy and the eponymous childhood education pedagogy, have several principles which help lay a foundation for learning, including:
- Children must have some control over the direction of their learning;
- Children must be able to learn through experiences of touching, moving, listening, and observing;
- Children have a relationship with other children and with material items in the world that they must be allowed to explore;
- Children must have endless ways and opportunities to express themselves.
Compare these principles with the formulaic definition of an Intelligent Agent:
- Agent: Anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators.
- Rational Agent: An agent that acts so as to maximize the expected value of a performance measure based on past experience and knowledge.
Clearly what is rational (or even the definition of learning) is much richer for Reggio Emilia children than it is for intelligent agents. The former has a community and a world to explore as a member, while the former simply acts on an environment in order to maximize expected value of a performance.
As noted, narrow AI is a refuge from the incalculable demands of Artificial General Intelligence, but too much has been cut out, and we end up with terrible tools. The clearest example of this is the hands off the wheel approach.
Amazon is a bit of a bipolar organization. For one, it prides itself on developing customer loyalty with cheap goods delivered quickly and consumer-friendly return policies. They are building lifelong customers and understand the value of that. On the other hand it is also humans who suffer at its misguided externalization of costs, including:
- Amazon contractors who are denied bathroom breaks, and Amazon actively discouraging unionization
- Amazon suppliers who are replaced with look-alike Amazon products, and rigged search results,
- Amazon has ridiculous order and receiving policies, and bot-driven support that drags on and on, sometimes for weeks, and where humans are unable to make necessary changes to a system that has been locked down.
It takes a while for a child to develop. Experiences, curated by parents, teachers, and childhood friends and family, as well as strangers. All of this requires good faith to make a well-adjusted and productive member of the community. Amazon is certainly not one to try and teach an Artificial General Intelligence, as it fails already with Narrow AI.