10 Useful Key Points I Got From Reading Impromptu

A.I. is probably the ultimate expression of our ingenuity as a species. We've learned to harness nature, dominate other species and create things like no other. However, we've always tried to figure out how to outsource the actual physical labor part, but why? So we can do more brain work. If we start to outsource brain work then what sorta use is that? That kinda defeats purpose...literally. He gets into it.

Like any tool, we can either use it to build up or destroy. The key is how we use the damn thing.

As A.I. and large language models improve, they will "require" less and less from us. When that's the case, we'll require more and more from ourselves. How we might do that is explored in the book (I found this fascinating).

You can only have zero risk and zero regulation in a world where there is zero progress. That's probably not a life worth living, yet... so many people living that life right now. A.I. can help make life better, and of course, it will require oversight and input from all manner of folks. Power that is democratized keeps it balanced.

Reid refers to us as Homo Techne - Technology apes. What is tech, but an external manifestation of our imagination? What is innovation, but a way to do something we couldn't do before using things we have in ways we never knew we could? It's like our species collective super power.

When we defer to things we sort of give up our agency, our individual power. It's the parent-child, slave-master, authority-subject relationship. People don't usually like to be an underling to someone else. However, resources are good and being able to access them quickly and easily is also considered good. Worth thinking about?

If you're not familiar with the turing test, it's a test to try and see if an artificial agent can convincingly fake it's way to getting a human participant to believe they're interacting with another real human. Nothing has been able to do that, but ChatGPT-4 seems the closest we've ever come. In fact, many folks have said if they didn't know they were dealing with A.I., they believe this would've fooled them, otherwise. I don't do this bit justice. More examples in the book. Unreal stuff.

In the book, Reid and Chat-GPT4 explore scenarios in which the success of A.I. either leads to a strange yet smooth "Algocracy" and or how it can lead to the dystopia most have some conceptual awareness of as a possibility (think 1984 or Brave New World stuff). The first principle thinking that goes into exploring this is pretty interesting stuff to ponder.

• Principle 1: Treat A.I. like an undergrad research assistant, not an omniscient oracle.
• Principle 2: Treat it like a director, not a carpenter. Directors tell the actor what they want, but don't decide how it gets done. Carpenters treat their tools as an extension of self. A.I. is not that type of tool. It's interactive but still does things it's own ways.
• Pinciple 3: Try it. Easier to just try and see what it comes up with than speculating what it might come up with. If nothing more, it helps stimulate your own thoughts on a thing.

Kills all manner of blank slate thinking in just about any field. It's... just read the darn book. so good.