"google maps is not an agent" and yet it does things like this bldgblog.com/2017/01/the-se
Conversation
maybe I'm just rephrasing you here, but tool AI and a company using it combine into an agent AI (albeit with a longer feedback loop)
2
1
no such thing as a real-world boxed AI; if you make maps that people use, you can expect them to impact the territory
1
1
so it seems less a choice between tool and agent, and more one between crippled ill-understood agent and better-understood agent
1
reason I'm so excited about this: a *very* similar thing crops up in a philosophy of magic(k) I'm collecting material for
1
where science is descriptive, concerned with models, while magic is about how those models form the world
Quote Tweet
Replying to @allgebrah and @sarahdoingthing
based on the descriptive/manipulative distinction; need a good magical counterpart to the experiment, maybe the REPL?
1
1
the scientific method works in a space where your predictions don't affect the outcome. that breaks down in the 'softer' sciences
1
which is why it wasn't physics which came up with the tinkerbell effect or the thomas theorem
2
a "magical method" as opposed to the scientific method would embrace its effects. more general, but less elegant & harder to manage
1
1
and lo behold, it is the same division I see between tool AIs (maps embedded in processes) and agent AIs (processes)
1
1
It's hard to talk about magic without sounding like a kook. Yet I think this is an interesting/fruitful way to go down