The AI effect as Heideggerian uncovering
In a previous post, I wrote about how concepts that have not yet been formed will save us from bad AI futures. In this post I give one articulation of what that might look like.
"AI is whatever hasn't been done yet."
- Larry Tesler, Xerox PARC.
The AI Effect
When arithmetic was still done by humans (whose occupation was ācomputerā) - and when the first automatic computer machines were being built - people looked to machines that could add numbers and perform computations on tables of data as āintelligent machinesā or āmechanical brainsā. Today, calculators or excel are not counted as AI, and we donāt find them very mysterious or existentially significant.
When Turing proposed that AI would one day be able to play good chess or converse like a human, he was met with lots of opposition. When computers finally did pass the Turing Test, it was not groundbreaking news. Everyone knew it was coming eventually after the major public launch of ChatGPT. Quickly, AI came to be mocked as being a āstochastic parrotā.
These phenomena have been called the AI effect. This is the idea that once a system is built that automates a new domain of human intellectual ability, it quickly loses its prestige and becomes ājust computingā. The next frontier then gets the name AI. For example, in the 1950s, search algorithms were considered AI, now they are just considered āsearchā.
Not weakness, but uncovering
I have always interpreted the AI effect as a weakness of people. I thought: weāre just failing to define terms precisely, and weāre failing to learn from history.
I have a new interpretation now.
What if we need to go through these processes of articulation of thinking and then its mechanisation, to even see other kinds of intelligence clearly.
After all, the best AI today thinks in human words. It is so dependent on our ability to uncover forms of thinking and mechanise them.
There is precedent for this, once farming went from being a majority to minority occupation, we invented so many other kinds of value.
Maybe all of human intelligence is being put to bad uses like: writing reports, doing desktop research, writing routine code, marking student homework, etc.
Maybe we will soon enter a golden age of not just artificial but also human intelligence. What if we will soon be freed to attend to others and to aesthetics, and maybe to politics and diplomacy and entrepreneurship - many things that we would rather do than what we do today.
What if humans are really a lot more powerful than we realise, but our potential is being capped by the amount of presently-understood intellectual load weāre having to carry, and the scaffolds (schools, financial incentives) we build to try to extract that intellectual labour from people. Once we can hand this off to AI, what if we find that we greatly change what we are attuned to and come to articulate new kinds of intelligence and new ways to navigate the AI-human relationship.