layout: single title: “Excel Can’t Run the Fed” tags: tech machine-learning
At lunch with some friends today, they asked me to define AI. Before going into the specifics of top-down approaches like expert systems or bottom-up approaches like machine learning, I told them the old joke in the AI community. It goes something like:
AI never really exists. If a computer system can accomplish a task, like play chess like Deep Blue or “understand” a human voice like Alexa, then that task is no longer artificial intelligence. But if a computer can’t do something like taking the dog for a run and then coming home safely, then accomplishing that task is still considered AI.
The question of what is “AI” versus what is just well-implemented current technology is complicated by this constant moving of goalposts.
Here’s an analogy. If you showed Microsoft Excel to people in 1920, most of them would consider its instantaneous management and calculation of complex numerical scenarios artificial intelligence. But no one who uses Excel today would generalize it’s amazing capabilities to thinking that it could act as Fed chair and manage the US economy!
People overgeneralize things they don’t understand. Since we can have useful interactions with the Amazon Echo, people immediately start thinking that C-3PO is just around the corner. This is really untrue.