Difference between revisions of "Reinforced Learning with Human Feedback"

From PKC
Jump to navigation Jump to search
Line 1: Line 1:
[[RLHF]] is also known as [[Reinforced Learning with Human Feedback]]. This is a testing or training strategy for data scientists to make sense or assign social meaning to data by human feedback. One way to think of [[RLHF]] at scale is to allow all [[human annotation]] and [[data editing history]] as a part of [[RLHF]]. This is particularly possible with a web-based interface that captures human inputs on portable networked devices. On top of that, if all these history are ordered and encoded with block numbers of some public [[blockchain]], the occurrence of all these [[RLHF]] activities could be universally integrated to reflect some statistical behavior of human intent as a collective. This is where [[accountability]] of individual and the collective can be operationally executed.
Reinforced Learning with Human Feedback can be abbreviated to [[RLHF]]. This is a testing or training approach for data scientists to use human feedback to make sense of or give [[social meaning]] to data. One approach to think of [[RLHF]] at scale is to include all human annotation and data editing history. This is especially true with a web-based interface that gathers human inputs on mobile networked devices. Furthermore, if all of this history is arranged and stored with block numbers of some public [[blockchain]]s, the incidence of all of these [[RLHF]] operations may be globally integrated to reflect some statistical behavior of collective human intent. This is where individual and collective [[accountability]] can be operationalized.  
    
    
<noinclude>
<noinclude>

Revision as of 06:43, 27 May 2023

Reinforced Learning with Human Feedback can be abbreviated to RLHF. This is a testing or training approach for data scientists to use human feedback to make sense of or give social meaning to data. One approach to think of RLHF at scale is to include all human annotation and data editing history. This is especially true with a web-based interface that gathers human inputs on mobile networked devices. Furthermore, if all of this history is arranged and stored with block numbers of some public blockchains, the incidence of all of these RLHF operations may be globally integrated to reflect some statistical behavior of collective human intent. This is where individual and collective accountability can be operationalized.


References


Related Pages


Retrieved from "https://pkc.pub/index.php?title=Reinforced_Learning_with_Human_Feedback&oldid=39064"