gosokkyu

エンド

  • 戦う人間発電所

owatte shimatta


ireneista
@ireneista

according to Reuters, among the issues in the WGA strike is the relationship of the profession to large language models, which some (rightly, in our view as a technologist) call "plagiarism machines"

this is not the only issue but it is certainly an important and timely issue. writers tend to be among the first to fully understand new things, it makes sense that they'd be striking over it. solidarity.


shel
@shel

A lot of people probably technically correctly will nitpick whether or not it's really plagiarism but you must remember the historical context. Much of what the current strike is about is streaming, which during the last strike was treated as an experimental new technology that amounted to five minute "webisodes" that couldn't amount to anything competitive with traditional TV. So the WGA made a lot of big concessions and now streaming has completely screwed over writers big time such that they're struggling to afford to live in the cities you have to live in to work in this industry.

This time around they're not taking any chances. There's a new technology. They're going to treat it as being the absolute biggest threat it could possibly be. If it could be used to replace a writer on their original show, or to replicate their tone, then even if it's not technically plagiarism it is still being used to eliminate union jobs. It might as well be a plagiarism machine.


mojilove
@mojilove

and the battle has already been lost to a great extent in the field of translation. I thought against writing this post because I wasn't sure if I could express it well, but then I saw the tweets below that do express it well:

A.I. (or machine translation) does not even need to produce very good writing, as its very presence is used to exploit labour even more than before. machine translation is at once seen as an all-mighty force that replaces human translators (i see the word "translation" being used very often as a synonym for "machine translation") - why pay a human a decent wage to do the job when the machine can get it done "instantly"? - but paradoxically, at the same time everyone1 knows that machine-translated text isn't good enough to be published as-is, so translators are hired to "edit" machine translation in a process that inevitably involves retranslating a majority of the text from scratch, only for less money because it is nominally "editing" work (this replaces the process of "editing" text that was first translated by a human translator who either did not have enough resources or who is not paid enough to care,2 and so it's cheaper as the first translator is taken out of the picture).

i wish translators would have had the bargaining power to resist this in the same way as the WGA is doing with this strike, and like the tweet above, i hope they win out


  1. well obv not everyone - the people who may very well be oblivious would likely not pay for a human translation in the first place

  2. i have heard of some people who do enjoy editing work but i personally find it immensely exhausting. editing requires (1) reading the original (2) reading the translation, (3) making sure they match adequately, (4) making changes (or rewriting), and (5) double-checking your own work. you can skip (2) and (3) by translating from scratch. in reality, i can edit human-translated text faster than i can translate from scratch, but at a cost of considerable psychic damage.


You must log in to comment.

in reply to @mojilove's post:

Translation industry is bleak. In addition to your points, I would add that the big companies pushing post-MT workflows already raced translation rates to the bottom without the need for ML/AI/new tech. The MT models are further underselling the work that, depending on language pairing and the company you're working for, may barely make any money as it is.