I think "we" needs some unpacking.
There's several stages in the training of an LLM. Early stages are the ones which use the massive data sets to teach it language and general facts about the world. Later stages use reinforcement learning to bake in the creator's preferences. That's where these kinds of pro-CCP biases would be deliberately introduced.
A "neutral" model would be one that has had those later stages abliterated. Whatever biases remain would be some kind of average of all the views of the people whose text is in the training corpus, which I hear is a pretty indiscriminate collection of literally anything people can download.
Jellyfin is pretty cool.
I always assumed this was astroturf
So are they going to do the NLRB election or no? Isn't that the normal process?
I like the new planks. Kinda bummed they're still shooting themselves in the foot with open borders and non-citizen voting.
Next up, cousin marriages to keep the family titles in the bloodline.
I wonder if any have started shipping with prepaid cellular
Maybe that's the context. All of my cloud hosting work is very large corporate systems in the big three clouds. You're talking about Hostinger or Scalahosting type services, yes?
All the cloud services I've used had virtual network firewalls for internal traffic. Security Groups for AWS, Cloud Firewall for GCP, etc. That's typically how lateral traffic gets managed. Defense in depth is good, but as I said I haven't used host firewalls in forever, and we get audited for security certificates every year, and no auditor has asked about them that I recall.
view more:
next ›