1.2 Autonomous AI?
While AI has up to now proved useful for automating certain kinds of tasks, there are concerns about how autonomous AI should be, particularly when it comes to making decisions that directly affect people. Indeed, a recent concern with how to make the design of AI more ethical, has in part focused on autonomy and the role of human oversight.
This ancient dream of creating intelligent servants (such as found in Homer) is seemingly quite directly reflected in the building of modern intelligent assistants, including a whole range of interactive technology such as Alexa, Siri and the like. It is still quite early days for such technology, and it remains quite rudimentary; moreover, there have been some serious mis-steps, such as the Tay chatbot created by Microsoft which was shut down in 2016, only 16 hours after it was launched on Twitter, due to its tweeting racist and sexist comments (The Guardian, 2016). More recently, a similar issue with the South Korean government’s Lee Luda led to it being removed from Facebook (McCurry, 2021). These are perhaps sensational examples of underlying deficiencies in the ethical design of AI. Furthermore, the goal of designing an intelligent assistant (or indeed ‘servant’), to carry out tasks in a relatively autonomous, albeit subservient, manner, remains problematic, particularly with respect to the risks inherent in it (some of which are exemplified by the Tay and Lee Luda chatbots mentioned above).
Sometimes quite startling connections between ancient and modern ideas can be discovered if you go searching in the right places. In this next activity you will reflect on the notion of artificial creatures that has persisted since at least Ancient Greece and see if there are such connections between historical and modern versions of such creatures.
Activity 1
- To begin, spend ten minutes searching the internet for one or two examples of artificial creatures throughout history.
- Then spend another ten minutes finding one or two examples of modern chatbots.
- Finally, compare your results to see if you can link the examples you found of ancient artificial creatures and modern AI systems in terms of their underlying purpose.
(Hint: Sometimes when trying to make such connections, you can start with the modern example and work backwards – so perhaps start with chatbots, think about their main purpose, and perhaps use this to guide your internet search.)
Discussion
- Some notable examples of chatbots include:
- ELIZA (Weizenbaum, 1966).
- A.L.I.C.E. (Chatbot.org, n.d.).
- If you find chatbots interesting, you might like to also look into the Loebner Prize (Wakefield, 2019).
- Note the main purpose of chatbots is essentially conversation. Maybe look for historical examples of machines which were used for simulating conversation.
- One possible historical example is the ‘talking head’ automaton that became popular in late medieval Europe, amidst rumours that pre-eminent scholars of the day, such as the English scientist Roger Bacon, and the German philosopher Albertus Magnus, owned such devices. These automata had specific capabilities more clearly associated with the heads of humans, e.g. they were used to simulate speech or could appear to return a person’s gaze.