Summary
Highlights
Humans have uniquely adapted to the environment not through genetic evolution like other species (e.g., mammoths adapting to Africa), but through technology. Approximately 60,000 years ago, humans migrated globally, not by waiting for genetic changes, but by creating tools and clothing (like mammoth fur) to adapt. Technology is the medium through which we transmit knowledge, dreams, and desires across generations, leading to increasingly advanced tools, from books to the internet.
A turning point arrived when machines began to resemble and learn from us, marking the birth of machine learning. This raises the concept of 'digital inheritance' in two ways. Firstly, it concerns the legacy behind AI development. Claude Shannon, the 'father of information,' defined information as 'a difference that makes a difference,' or something that causes a change in behavior, using the analogy of Maxwell's demon. He demonstrated this with a mechanical mouse in a wired labyrinth that learned to find its way out through trial and error, a precursor to modern robots like the Roomba.
Norbert Wiener formalized the idea of using information to control machines, coining the term 'cybernetics.' This involved a sensor between the machine and reality, translating information to guide machine behavior, like automatic doors. Margaret Mead extended this concept, suggesting that if machines could be controlled through feedback, so could human and social processes, leading to system theory. Wiener, however, raised a critical question in 1967: if the most effective sensor to control machines is human, then what will be the human use of human beings in a cybernetic society?
Cybernetics was a proposed name for artificial intelligence, but John McCarthy chose 'artificial intelligence' partly due to disdain for Wiener. Wiener's writings, however, influenced Martin Heidegger, who, in a conference titled 'The End of Thought in the Era of Philosophy,' argued that if we can control reality through information and machines, we lose the need to truly understand it. Secondly, he questioned whether humans control machines or machines control humans, citing the example of smartphones: do users control the display, or do notifications control the user? This highlights the inherent control dimension in AI.
The power of these machines is evident in cases like AI-generated images (e.g., the Pope in a Balenciaga jacket). These technologies, amplified by vast amounts of data, lead to foundational models like large language models (LLMs) such as ChatGPT. ChatGPT creates a broad map of reality from data, identifying patterns. While powerful, ChatGPT can 'hallucinate' or produce incorrect information. However, its new version can also interpret images and assign meaning, raising concerns about machines mediating meaning rather than people. There's also the risk of 'visual hacking' where AI can be tricked into misinterpreting images.
Given the 'inheritance of control' embedded in these machines and the cultural inheritance they transmit, we must ask: what do these machines leave for us and future generations? Digital services often replicate analog processes but can also create entirely new cultures. To ensure AI enhances humanity rather than diminishes it, we need 'algorethics' – a new chapter in human ethics that is both comprehensible to humans and executable by machines. This means establishing guardrails to steer AI development toward paths beneficial for humanity, moving away from an inheritance of war or control towards a more ethical future.