Harper's Magazine - The Hoffman Wobble, or how Wikipedia infected out minds

[suckadmin]

It is otherwise a work to respect. It correctly identifies Wikipedia's fate is already sealed, hastened with the approach of AI. These men who stare at goats, they "represent the end of Wikipedia."

Doubtful. An "AI" encyclopedia would possess none of the social pretenses that made Wikipedia an effective tool for social engineering. The public would (rightfully) distrust it.

Yes because it's pretty much guaranteed that any of the available LLMs are already trained on material from WP as well as other even less reliable sources. So the result is going to have all the same biases as that material

Even if somebody was able to take the time and energy ( both human and GPU) to properly train from direct sources they would still have the issue of copyright infringements because they would need to name their sources to have any credibility. The AI would still hallucinate and the closer you get to contemporary times and dominance of the internet the less reliable online resources probably become or the more likely certain bits of information would trend and become prevalent in the training regardless of their lack of truth or actual relevance.

Interestingly unlike other marketing trends like sticking HD or 4K on products that have little to do with video using AI as a marketing buzzword is failing

https://www.techradar.com/computing/art ... -marketing

[Archer]

For your amusement, I passed this along to parish coroner Bones, who has now published his examination in the disinformation report.

Gee, thanks a whole lot.....Smallboner is a case study of lack of self-awareness.

But first a warning: to address the serious questions raised by the Harper's article (more properly titled The Hofmann Wobble – Wikipedia and the problem of historical memory), which is a combination of fact and fiction in the form known as autofiction, I try my hand for the first time at literary criticism.

Ha ha. Give it up.

And at least one sucker didn't "get" it.....

After reading Lerner's "fiction" I find myself frustrated. This sort of behavior does the project – and the trust within – no good. Even this weasel confession is lacking any awareness of the harm done. Ckoerner (talk) 20:27, 5 December 2023 (UTC)

IT'S FICTION, YOU GODDAMNED IDIOT. Are you trying to claim that ANY discussion of WP content abuse is "bad for the project"? FOOL. People are abusing WP with sock accounts RIGHT NOW and CONSTANTLY, they have done so since it started nearly 23 years ago, and they will not stop. Are you another Wiki-Fuck cult victim who wants to "rewrite history" and censor everything to make Wikipedia look "perfect"?

And you're a Wikimedia employee. Do you realize how stupid this makes the WMF look? Are you even less self-aware than Smallboner?

Just BTW: idiots are already fooling with Lerner's BLP.
https://en.wikipedia.org/w/index.php?ti ... 1188583172

"The argument that to tell the truth would be ‘inopportune’ or would ‘play into the hands of’ somebody or other is felt to be unanswerable, and few people are bothered by the prospect of the lies which they condone getting out of the newspapers and into the history books. [...] A totalitarian society which succeeded in perpetuating itself would probably set up a schizophrenic system of thought, in which the laws of common sense held good in everyday life and in certain exact sciences, but could be disregarded by the politician, the historian, and the sociologist. " -Orwell

[ericbarbour]

This is how you know Wikipedians want to shit on someone's head: when these idiots show up.

curprev 16:27, 30 November 2023‎ Smallbones talk contribs‎ 19,657 bytes −36‎ direct link to site undo
curprev 15:30, 29 November 2023‎ Fuzheado talk contribs‎ 19,693 bytes +27‎ add ref undo Tag: Visual edit
curprev 15:29, 29 November 2023‎ Fuzheado talk contribs‎ 19,666 bytes +208‎ →‎Awards: add ref for Pulitzer undo Tag: Visual edit
curprev 20:59, 28 November 2023‎ Drmies talk contribs‎ 19,458 bytes −1,328‎ not a resume: rm individual short stories undo

[Archer]

This is how you know Wikipedians want to shit on someone's head: when these idiots show up.

curprev 16:27, 30 November 2023‎ Smallbones talk contribs‎ 19,657 bytes −36‎ direct link to site undo
curprev 15:30, 29 November 2023‎ Fuzheado talk contribs‎ 19,693 bytes +27‎ add ref undo Tag: Visual edit
curprev 15:29, 29 November 2023‎ Fuzheado talk contribs‎ 19,666 bytes +208‎ →‎Awards: add ref for Pulitzer undo Tag: Visual edit
curprev 20:59, 28 November 2023‎ Drmies talk contribs‎ 19,458 bytes −1,328‎ not a resume: rm individual short stories undo

His BLP doesn't seem much worse for wear. In the very same essay I quoted from earlier, Orwell also wrote the following - "There is a whole series of converging reasons why it is somewhat easier for a poet than a prose writer to feel at home in an authoritarian society. To begin with, bureaucrats and other ‘practical’ men usually despise the poet too deeply to be much interested in what he is saying. Secondly, what the poet is saying — that is, what his poem ‘means’ if translated into prose — is relatively unimportant, even to himself."

I find it difficult to read Lerner's piece. If one has a serious critique to make then one should use plain language, not a pretentious and inhumanly long narrative.

[Archer]

I should add that my comments above aren't merely the petty nit-picking of another critic (if Lerner could be called a critic). Narrative is perhaps the most common mode of expression in marketing and propaganda. It seems to me that people frequently adopt this manner of writing and speaking; presumably it is impressed upon the public through constant exposure. Yet this is not a direct way of speaking, nor is it a natural way to express critique or participate in discourse. Harpers gets away with it upon vague literary pretenses. Personally I do my best to keep it out of my own language. If one has a real, substantive point to make, then there's no reason to use parable or storytelling.

Just think how much more compact and precise Lerner's article could be if it were written as a serious commentary. Of course this would require more serious consideration - one would need to formulate a specific point or hypothesis and build an argument to support it. Yet this is how one should address a serious subject.

I dislike Harpers. They enshrine bullshitting as an art-form. It's not a normal way to talk.

[boredbird]

Even if somebody was able to take the time and energy ( both human and GPU) to properly train from direct sources they would still have the issue of copyright infringements because they would need to name their sources to have any credibility.

It seems you've stumbled across the not-so-secret link between plagiarism and IP theft. "It isn't theft because we didn't know!" Not limited to big tech.

AI has innumerable potential applications, but from Silicon Valley's perspective, plausible deniability is the "killer app", as they call them.

[suckadmin]

Even if somebody was able to take the time and energy ( both human and GPU) to properly train from direct sources they would still have the issue of copyright infringements because they would need to name their sources to have any credibility.

It seems you've stumbled across the not-so-secret link between plagiarism and IP theft. "It isn't theft because we didn't know!" Not limited to big tech.

AI has innumerable potential applications, but from Silicon Valley's perspective, plausible deniability is the "killer app", as they call them.

Ignorance of the law is no excuse.. unless of course you're tech innovator.

It's like that line from Jurrasic Park

"Yeah, yeah, but your scientists were so preoccupied with whether or not they could that they didn't stop to think if they should."

It was reasonably fair use when it was all just research but once it became viable for commercial exploitation the shit hit the fan..

On the upside or maybe the downside we may start building more nuclear power plants just to power the rendering of photorealistic five titted anime girls

[suckadmin]

I went to look up the status of any one of the various class actions against ai developers since they will determine the lay of the land and look here's new one!

https://www.hollywoodreporter.com/busin ... 235968822/

my prediction is that even if the past models that used unpermitted web scrapings get restricted that new models from licensed stock libraries and fine print licensing when combined with end user training will result in the same situation. Also the resources to do the training will probably decrease significantly making it easier and more affordable to do said training.

Seems like all the billions they're spending on it is not going to reap huge rewards like they were hoping

[Archer]

my prediction is that even if the past models that used unpermitted web scrapings get restricted that new models from licensed stock libraries and fine print licensing when combined with end user training will result in the same situation. Also the resources to do the training will probably decrease significantly making it easier and more affordable to do said training.

Indeed, the point is somewhat lost unless they publish the architecture, training process, and training set so that the public can reproduce their results. I'm not inclined to take their word for it. Though it's likely there are various black projects and corporate schemes involving LLMs and will not be made public.