You could of course have a text editor which, after each sentence, prompts you with "could you state this in the form of a bet?" and gives a standard bet format -- "by [time], we will/will-not observe [foo], conditional on [bar], as measured by [third-party source of info]"
-
-
Show this thread
-
for example, if you asked me to restate the past tweet in the form of an empirical prediction: "by 2022, we will not observe an automated rewriting tool like this, conditional on at least $1M of person-hours being devoted to it, as measured by a Google search for such tools"
Show this thread -
Once you have a prediction, of course, you can lay odds on it, and afaik there's already adequate software for setting odds and finding someone to take the other side of bets.
Show this thread
End of conversation
New conversation -
-
-
I've built prediction tools inspired by
@PTetlock's work. Compiling predictions like this is a tricky, human-in-the-loop bottleneck, always. -
oh? can you link or describe the tools?
- 1 more reply
New conversation -
-
-
I think this is an important problem though. At the moment it's easy for an AI to do to things: 1) write fluent text about arbitrary events in the real world 2) make quite accurate probabilistic predictions in toy worlds like StarCraft or limited domains like face detection
-
But modern AI is totally useless at 3) making reasonable predictions about events that are not in a limited or toy domain AND it also cannot 4) take a statement in natural language and relate that statement to confirmatory or disconfirmatory evidence
- 4 more replies
New conversation -
-
-
I think it *cannot* be automated (short of an AGI), due to "the cluster structure of thingspace" and related issues.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.