For the past daysI have been unable to update my system because of Pacman issues, so i have disabled the Pacman repository and executed:
sudo zypper dup --allow-vendor-change, then I have the following unresolvable issue"
t]
Retrieving: ghc-containers-0.7-1.3.x86_64.rpm .....................................................[not found]
Retrieving: ghc-containers-0.7-1.3.x86_64.rpm .....................................................[not found]
Retrieving: ghc-containers-0.7-1.3.x86_64.rpm .....................................................[not found]
Retrieving: ghc-containers-0.7-1.3.x86_64.rpm ...........................................[error (185.8 KiB/s)]
Failed to provide Package ghc-containers-0.7-1.3.x86_64 (Main Repository (OSS)). Do you want to retry retrieval?
[download.opensuse.org-oss|http://download.opensuse.org/tumbleweed/repo/oss/] Can't provide file './x86_64/ghc-containers-0.7-1.3.x86_64.rpm' from repository 'download.opensuse.org-oss'
History:
- Downloaded data exceeded the expected filesize '636.2 KiB' of 'http://cdn.opensuse.org/tumbleweed/repo/oss/x86_64/ghc-containers-0.7-1.3.x86_64.rpm'.
- Can't provide ./x86_64/ghc-containers-0.7-1.3.x86_64.rpm
Abort, retry, ignore? [a/r/i] (a):
Please, next time use the spelling “Packman” when you mean Packman.
“Pacman” is also name within the Linux world. It is the package manager of Arch Linux. As such it is also mentioned every now and then in these forums. So just to avoid confusion.
@hcvv, you will get a kick out of this. When I was looking up “packman” at a few different sites, the AI assistant started with a monologue talking about “pacman” and used the the wrong spelling throughout the article even though it was talking about the functioning of “packman”.
What is even more worrisome to me is how do corrections to glaring wrongs ever get made. There’s no feedback so that corrections can get back into the learning and model training that I am aware of.
I worked on and applied AI applications 30+ years ago before it became de rigueur. For our app, the models got retrained continuously.
How should this work with the public models? They get fed with gazillion sets of data. To filter out wrong data, a human would need to filter the gazillions sets of training data. Impossible. Instead, if not carefully selected, AI will train on its output. That is a downward spiral. AI is doomed to degenerate quality wise over time…there are intersting studies and articles about this fact.
I agree it is really a slippery slope. There is no way to get any corrective data back into the training. Who is even tasked to do that and there is no lists of corrections that is stored anywhere. People just fix the AI wrongs on their own and move on.