I went back through a few months of comments and restored the ones Google wrongly marked as spam. There were a lot of them, including some Mr. Hines made multiple times, along with a comment that he couldn’t understand why they kept disappearing.
The effect of AI is likely to include increased speech suppression. Disapproved ideas will be much harder to express because human attention will no longer be a limiting factor on how much suppression can be done.
12 comments:
It seemed to me that you'd cleared that up at the time, or at least successfully identified the problem. Has Alphabet (Google's owner) been at it again
Eric Hines
They must have all been marked as spam again. Unlike the AI, my time and attention is limited; they’re eventually going to be able to clear up whatever thought they dislike.
Reminds me of Heinlein's The Puppet Masters, only they're not alien parasites, which Heinlein was analogizing with then communist Russia, but human AI programmers, who have the same goal as those communist Russians.
Eric Hines
You've probably considered moving to a non-Google platform. Cost is a factor, of course, and I imagine migrating everything takes some time. Were there other reasons for not moving?
I'm in the process of de-Googling my life at the moment. I'm about to switch to Graphene or a modified Lineage OS version of Android, moving back to Linux (I reluctantly went w/ Windows machines for work, but that job has ended), & finding & learning alternatives for the rest of Google.
One "interesting" thing about the Google spam/thought control is that it doesn't necessarily zap comments right away. Comments can be visible for several days before Google pounces.
Yeah, that's right. I noticed the comments on the Red Flag of Revenge post had decreased to 13, which is what caused me to go and look. That was from Wednesday, and I'd already reset the comments on it once. Their machine came back around and hit it again.
There never were comments regardless of your findings.
(Signed) Google.
There seems to be a general problem online with The Disappearance of the Past. Any link to a media site older than a few years is likely not to work...usually, the article is still there, but the structure of the site was changed such that the URLs changed and the previous links no longer work. At Chicago Boyz, we are encountering a Wordpress phenomenon such that older posts display with weird and unpleasant typography that makes them difficult to read. (Any WP experts who could help us with this?)
At Quillette, I noticed that a post I wrote a couple of years ago is still there, but all the comments have disappeared. Don't know if problem is specific to this post or is generic--submitted a support query.
That's all in addition to the explicit censorship algorithms and humans.
(Particularly ludicrous example of algorithmic enforcement of Correct Information at FB: at a discussion group for pilots and air traffic controllers, a pilot was thanking controllers at a particular airport for expedited handling of an emergency medical flight, The algorithm noticed the word 'medical' and added a cautionary notice: "For accurate medical information, see these (official) sites".
Couple things:
a general problem online with The Disappearance of the Past.
Some (much?) of that disappearance is deliberate by the press, among other outlets, as they rewrite their own histories and try to pretend the actual history never happened, or they alter what they wrote and try to pretend that's what they'd written originally. Sometimes they'll append to the latter their claim to have "updated" or "corrected" the original.
The algorithm noticed....
No, the humans who programmed the algorithm programmed it to "notice."
Eric Hines
Eric H...sure, the algorithm was programmed, most likely, any word or phrase such as 'medical' or 'healthcare' in a post would trigger the routine to insert the 'here's where to get real information' block of text. But 'the algorithm noticed' seems like a reasonable way to describe how the algorithm responded to the post, even though it (the algorithm) is not conscious and can't really 'notice' anything.
Soon if not quite yet, it may work differently. A Large Language Model may be fed a large # of posts and of human evaluations of those posts, and then the LLM would generate the code that decides which posts to flag based on optimizing closeness to what an actual human would do. And that code would likely be pretty incomprehensible and it would generally not be possible to get the WHY as to why a particular post was disapproved.
A Large Language Model may be fed a large # of posts and of human evaluations of those posts, and then the LLM would generate the code that decides which posts to flag based....
Sure. But it's still human programmed, including how the algorithm should subsequently carry out its own code writing. I can train rats to go through a series of apparently unrelated tasks of varying complexity in order to get its reward, too. Was the rat acting independently, or was it operating IAW my instructions? I think that's pretty clear.
The incomprehensibility of the algorithm's code isn't particularly important; it's still just doing what the human programmers told it to do--including how to evaluate all those posts, in ways closely analogous to the way my rats evaluated their task sequences IAW how I trained them to do so.
Shorthand descriptions of what the algorithms seem to be doing are fine, but only so long as the origins of those decision criteria and evaluation methodologies are kept in mind. Without that clarity, we get, in very short order, to "guns kill," losing track of the fact that, no, it's the gun operator who's doing the killing.
Eric Hines
This is teaching people a good lesson. If I recall, they shadow banned my wordpress blog since like 2007 or 2009, to 2023. A solar flare suddenly unshadowbanned me in march. Until then, searching ymarsakar would just bring up my blogspot old blog, but not the wordpress or other stuff. These are all part of the "process" so to speak.
Post a Comment