"Copy what you hear" is good advice outside of ham radio, as well, but it does
require careful listening.
My company is getting paid to have me do a webinar on the risks of generative
AI, such as OpenAI ChatGPT that everyone is playing with.
Many of those AI programs use what is called a Large Language Model. That
model, and others, have been shown to suffer what has been given an odd name -
Model Autophagy Disorder, which sounds like it would only inflect those skinny
fashion runway walkers...
Basically, if those AI models "eat their own dog food" and ingest their own
creations as part of continuous updating of the model, then errors get
propagated and increase the likelihood of future errors very quickly.
It is kind of like a bad RBN or cluster spot for BY5IJ getting worked and then
respotted by someone S&Ping with a loaded available Mult/Q window and "Spot all
S&P" QSOs - with the added impact as if BY5IJ call was also then added to the
SuperCheck partial database and all history files. And all that happening at a
global scale.
There are known solutions to the problems, we are lucky in contesting in that
log checking will get you if don't focus on "hearing before copying" and just
rely on spots/history files, etc. and the consequences are mostly focused on
you and far from life threatening!
73 John K3TN
_______________________________________________
CQ-Contest mailing list
CQ-Contest@contesting.com
http://lists.contesting.com/mailman/listinfo/cq-contest
|