Re: Can AI make a better Marathon?


	/-/S'pht-Translator-Active/-/

Re: Can AI make a better Marathon?

Posted By: Document Date: 2/12/26 11:02 a.m.
In Response To: Can AI make a better Marathon? (ehyto)

With no actual testing I'd say: "know", no. "Be able to find", probably, given adequate problem definition and iteration loop, and a smart enough model running it.
Comments I was coincidentally seeing on Twitter recently:
"verifiability is magic - not sure what verifiable problems can’t be solved with AI"
"You must, at any cost, remove yourself as a blocker from the core evaluation loop... You're never going to achieve takeoff hand validating"
But I've never tried any of this myself.

[ Post a Reply | Message Index | Read Prev Msg ]
Pre-2004 Posts

Replies:

Can AI make a better Marathon? ehyto 2/12/26 6:37 a.m.

Re: Can AI make a better Marathon? Document 2/12/26 11:02 a.m.

[ Post a Reply | Message Index | Read Prev Msg ]
Pre-2004 Posts

Your Name:
Your E-Mail Address:
Subject:
Message:
: With no actual testing I'd say: "know", no. "Be able to : find", probably, given adequate problem definition and iteration : loop, and a smart enough model running it. : Comments I was coincidentally seeing on Twitter recently: "verifiability : is magic - not sure what verifiable problems can’t be solved with AI" : "You must, at any cost, remove yourself as a blocker from the core : evaluation loop... You're never going to achieve takeoff hand : validating" : But I've never tried any of this myself.

If you'd like to include a link to another page with your message,
please provide both the URL address and the title of the page:
Optional Link URL:
Optional Link Title:

If necessary, enter your password below:
Password:

Problems? Suggestions? Comments? Email maintainer@bungie.org
Marathon's Story Forum is maintained with WebBBS 5.12.

	Can AI make a better Marathon?	ehyto	2/12/26 6:37 a.m.
	Re: Can AI make a better Marathon?	Document	2/12/26 11:02 a.m.



	Your Name:
	Your E-Mail Address:
	Subject:
	Message: : With no actual testing I'd say: "know", no. "Be able to : find", probably, given adequate problem definition and iteration : loop, and a smart enough model running it. : Comments I was coincidentally seeing on Twitter recently: "verifiability : is magic - not sure what verifiable problems can’t be solved with AI" : "You must, at any cost, remove yourself as a blocker from the core : evaluation loop... You're never going to achieve takeoff hand : validating" : But I've never tried any of this myself.
	If you'd like to include a link to another page with your message, please provide both the URL address and the title of the page:
	Optional Link URL:
	Optional Link Title:
	If necessary, enter your password below:
	Password: