Human Roles in Growing AI

Human Roles in Growing AI

Human Roles in Growing AI

Human Roles in Growing AI

I would rather see intelligence as an ability to grow in knowledge and wisdom than as a product made by genius engineers that is ready for sale. If we really wish AI to own the same intellectual ability as us, we should be aware of the process of how we develop our intelligence. We, as intelligent beings, are born as a blank sheet of paper. To acquire intellectual growing, children need to learn from the environment, parents, teachers, friends and even opponents except engineers, that is, children need to learn from interaction with other humans. From this point of view, changing children with AI, I am curious about what roles humans could play in the long and continuous process of growing AI except acting as engineers and scientists to built AI. Therefore, to grow AI may be a much more challenging task than to build one, requiring all humankind involved to become an indispensible part of the whole AI learning ecosystem.

I can think of several roles that humans can take listed as follows:

  • Teachers. This role should provide clear and correct answers for AI as in supervised learning. The answers could be a right final outcome to be predicted such as labels, or a sequence of correct actions that let AI be able to mimic step by step. More flexibly, instead of final or all-the-way supervising, teachers can pick the critical moments to give supervison more efficiently.
  • Masters. Unlike teachers, masters do not have to hold enough domain knowledge to teach machines exactly what to do. Instead, they should be able to sense whether an action or a produced result is acceptable or not, and provide appropriate encouragement or rewards as well as warnings or penalties as in reinforcement learning. This kind of relationship between humans and machines is like the one between masters and pets. Compared with teachers, the master role requires less human’s effort, but with more people involved in the direction interaction with machines, it can release the burden on AI engineers and provide much richer guidence from the human society.
  • Opponents or competitors. From its literal meaning, this role plays against machines, a sort of adversarial training to force machines to make progress on its own. Playing chess or go between humans and machines belongs to this domain, but it demands all-the-way sparring and may consume too much time from humans. Another type of opponents is a trouble maker who can creat obstacles for machines from time to time.
  • Parteners or cooperators. Contrary to opponents, this role requires huamns to work with or aid machines to achieve some goal. This sort of aid is incomparable with that of teachers or masters but no more than a communication channel or a helpful behavior, as human parteners are also a game player similar to machines and may even know little about this world. Even so, human parteners can still give extra observed information to machines and watch how machines react, which provides an opportunity for both to communicate with and learn from each other.
  • Agent builders. This role is actually played by engineers. An agent builder can decide how to design and assemble an AI machine, such as what types of sensors it has, what actions it can take, and how deep its neural network can go.
  • World builders. This role acts as a super administrator who owns the hightest priviledge to add, remove and modify every detail in this world, and to design new sceniors and set up new tasks to train machines.