@rachel What’s wrong with “delve”? It’s a perfectly fine and useful word. But I’m not a native speaker of English, so what do I know? The idea that the knowledge of the English language I acquired during 40 years is “inhuman” is not very reassuring.
There is hair that needs some pulling out but it is not OPs.
@chris @rachel > What’s wrong with “delve”? It’s a perfectly fine and useful word. But I’m not a native speaker of English, so what do I know?
Just by using “delve” regularly as a non-native speaker, it shows you recognize quite a bit more than the native English-speaking techbros behind the push for AI everywhere.
There’s evidence that ChatGPT’s fondness for “delve” (and likely other English words) comes in large part from people in places like Nigeria where (a) “delve” is common in the local flavor of English and (b) a lot of people have contributed to LLM model training and refinement.
@dpnash @rachel Thanks, that’s an interesting perspective and a revealing incident. Language evolves, and as with biology, the result varies for differing habitats. Since “#AI” that is based on #LLM is frozen in time by the corpus it was trained on, it only reproduces but never explores or invents. LLM are per definition conservative. (1/3)
“AI” marking #delve as evidence for cheating with “AI” is a strong reminder that “AI” still is nothing but “Autocorrect on steroids”.
Using “AI” to categorize human activity is a way of pressuring humans to behave in the way that has been defined by the corpus the LLM has been trained on. The people subjected to this process have no say in the choice of corpus or in the training of the black box that will be used on them. (2/3)
Using “AI” to coerce humans to behave according to digitized expectations is autocorrect for human behaviour, it is classist, it is imperialist, it is a form of violence. (3/3)
If you don't already follow him, mekkaokereke@hachyderm.io has had some enlightening things to say about Nigerian-style English and its interactions with AI and such.
Generative AI is making our language and communication poorer. It does this not just by outputting its own garbage but also because now we have AI making arbitrary determinations of what words and phrases we *humans* are or aren't allowed to use.
@heydon I read this just two days after having a discussion with people about an AI-generated text about our work group which was of the most generic kind you tune out and skip while reading. Despite me mentioning that, it was lauded as a nice piece if work. This has something to do with having a text, picture or whatnot that has no other task than just being there. Visual or textual noise. Anything slightly individual or deviating from this would draw criticism because " thats not...
Rachel
in reply to Artemis • • •in some convo at work, it was said that "delve" was a nearly perfectly indicator
Ok so I know entire categories of nerds who would write like that right?
chris@strafpla.net
in reply to Rachel • • •@rachel What’s wrong with “delve”? It’s a perfectly fine and useful word.
But I’m not a native speaker of English, so what do I know?
The idea that the knowledge of the English language I acquired during 40 years is “inhuman” is not very reassuring.
There is hair that needs some pulling out but it is not OPs.
David Nash
in reply to chris@strafpla.net • • •@chris @rachel
> What’s wrong with “delve”? It’s a perfectly fine and useful word.
But I’m not a native speaker of English, so what do I know?
Just by using “delve” regularly as a non-native speaker, it shows you recognize quite a bit more than the native English-speaking techbros behind the push for AI everywhere.
There’s evidence that ChatGPT’s fondness for “delve” (and likely other English words) comes in large part from people in places like Nigeria where (a) “delve” is common in the local flavor of English and (b) a lot of people have contributed to LLM model training and refinement.
premiumtimesng.com/opinion/689…
DELVE: Artificial intelligence, diversity and inclusivity, By Chiamaka Okafor
Chiamaka Okafor (Premium Times)chris@strafpla.net reshared this.
chris@strafpla.net
in reply to David Nash • • •Language evolves, and as with biology, the result varies for differing habitats. Since “#AI” that is based on #LLM is frozen in time by the corpus it was trained on, it only reproduces but never explores or invents. LLM are per definition conservative. (1/3)
chris@strafpla.net
in reply to chris@strafpla.net • • •“AI” marking #delve as evidence for cheating with “AI” is a strong reminder that “AI” still is nothing but “Autocorrect on steroids”.
Using “AI” to categorize human activity is a way of pressuring humans to behave in the way that has been defined by the corpus the LLM has been trained on.
The people subjected to this process have no say in the choice of corpus or in the training of the black box that will be used on them. (2/3)
chris@strafpla.net
in reply to chris@strafpla.net • • •JimmyChezPants
in reply to chris@strafpla.net • • •@chris
If you don't already follow him, mekkaokereke@hachyderm.io has had some enlightening things to say about Nigerian-style English and its interactions with AI and such.
And a lot of other stuff too of course.
chris@strafpla.net
in reply to JimmyChezPants • • •Artemis
in reply to Artemis • • •Artemis
in reply to Artemis • • •Tell you what: "devoid" was definitely a word I knew and used in elementary school.
Am I normal? Nah, not really.
Am I human? I'm pretty damn sure I am, yeah.
Artemis
in reply to Artemis • • •FoolishOwl
in reply to Artemis • • •Large Heydon Collider
in reply to FoolishOwl • • •ManniCalavera
in reply to Large Heydon Collider • • •This has something to do with having a text, picture or whatnot that has no other task than just being there. Visual or textual noise. Anything slightly individual or deviating from this would draw criticism because " thats not...