I'm wondering if anyone else is feeling like they serve as lightning rod for this kind of behavior. Maybe @olivia.science or @irisvanrooij.bsky.social ?
11.03.2026 13:23
👍 14
🔁 0
💬 1
📌 0
I'm wondering if anyone else is feeling like they serve as lightning rod for this kind of behavior. Maybe @olivia.science or @irisvanrooij.bsky.social ?
So rather than listening to the doubt, and confronting those possible harms, they seek to erase it, by going to the source, as it were.
>>
... and they have identified that doubt with something connected to my public scholarship (often just the phrase "stochastic parrots").
>>
I'm guessing/hoping that there actually remains a seed of doubt in those who are most loudly proclaiming the "usefulness" of this technology -- a doubt that they might be harming things they care about (however larger their circle of concern is),
>>
So why is it so important to convince me in particular to say the thing they want to hear?
>>
And anyway most of the people who seem to be seeking this validation from me aren't specifically profiting from "AI" that I can tell.
>>
I doubt it's because they fear whatever meager power I have in this situation -- I mean, yes, I am trying my best to bring about better circumstances by helping both ordinary people & policymakers understand the hype & harms and make better decisions -- but I don't control funding.
>>
This is different from arguing with me (or someone else holding my position) in that the seem to have a need to get me to agree with them.
>>
This is part of a pattern I've started to notice: There are people out there who are loudly committed to off-loading their work to chatbots and come seeking my validation for their position.
>>
Oh excellent! I had remembered slop as a combining form being discussed but not sloppelganger itself.
Is it too early to submit nominations for 2026 WoTY @americandialect.org ?
Is it too early to submit nominations for 2026 WoTY @americandialect.org ?
lmaooooo sloppelganger
I'm not sure why you feel so entitled to my time and attention and I am not going to retype all of my arguments here 300 characters at at time for you, especially since you are engaging with none of them. I'll leave you with this, on automatic transcription:
youtu.be/LfOiEV_x1fg?...
Today's mansplainer award goes to the prof in (checks notes) a School of Global Affairs, who seems to believe that he knows better than I do about how language technology works.
So you're really saying "I really want to hear
@emilymbender.bsky.social say "The fascist technology is a remarkable achievement with all kinds of impressive use cases like translation [even though I disagree with how it's done]."
... which, no.
Meanwhile, @akoustov.bsky.social you seem to have read this post very differently from how I do, skipping right over the fascism bit.
bsky.app/profile/chag...
>>
Language models based on the transformer architecture can be used in ways that are beneficial, but such beneficial use requires transparency into training data (as I have been arguing since 2018) and that doesn't even touch the unethical production of these systems.
>>
That seems like such an odd request, honestly.
Let me state clearly (as I have many times before) that I see no beneficial use cases for synthetic text.
>>
I'm not even talking about norms, but simple affordances of the platform. If you quote post, it goes to your followers. If you reply, then someone who saw my post that you were replying to (i.e. my followers) is more likely to see it.
p.s. When you say you agree with all three, do you mean you agree with me that these are bad, or that you are in fact promoting all three?
bsky.app/profile/akou...
If you were honestly trying to reach my followers, you would have replied, rather than quote-posted --- so I don't take you to be engaging in good faith here.
Nonetheless, I will reply:
As for this, no thank you. But I do hope someone else has the time to look at your unverified translations and find all the howlers for you.
bsky.app/profile/akou...
What other resources might we think about providing to asylum seekers, if we really cared?
For ex:
respondcrisistranslation.org
>>
As for your imagined use case: have you evaluated how well these systems work for translating from English into the language of the asylum seekers you have in mind? Is the output comprehensible? Accurate?
>>
Meanwhile, the ersatz flueny of LLMs makes the output look *too* smooth, leading users to misjudge whether or not translations are accurate.
>>
However, that doesn't mean it is a good or reliable or sufficiently transparent way to to MT. A purpose built system, trained and evaluated on MT specifically, is a better choice.
>>
It is also true that the chatbot providers have worked on making MT accessible via the chat interface.
>>
Yes, language models are an important component of machine translation systems, and yes the transformer architecture in particular and so-called neural models in general have led to improvements in MT.
>>
If you were honestly trying to reach my followers, you would have replied, rather than quote-posted --- so I don't take you to be engaging in good faith here.
Nonetheless, I will reply: