Yikes. Once again, the Ethics Alarms attic is chaos, and I am waaaay behind in covering important ethics stories, breaking ethics stories and developments in recent ethics stories I did get around to. Yesterday, for example, we learned that LA Dodger two-way superstar Shohei Ohtani’s good friend and interpreter stole 16 million bucks from the player to cover his illegal gambling problem, not “just” four million.
I’m hoping the Wisdom of Crowds can help clear the metaphorical decks today.
Ethics Alarms 
Have you seen the recent controversy on Twitter (I will never call it X) between Paul Graham (Founder of Y combinator) and Nigerian Twitter concerning the use of English and diction?
PG identified some words like “delve” and “burgeon” as indicators that someone relied on Artificial Intelligence in writing a piece. When Nigerians, and indeed persons from many other Commonwealth countries tried to inform him that those words are normal words for us by reason of our education, our native languages which tend towards the expressive and dramatic, use in everyday life, colonial history, and the means by which we were introduced to the English language – books, he doubled down. When I showed him a screenshot of my WhatsApp communication a few days earlier where I used the word “burgeon” in colloquial context, he blocked me.
Many others spoke in his defence about the American preference for informal language even in formal settings, and that the use of those apparently arcane words would be frowned upon and even considered grandiloquent (“showing off” they called it).
I think it speaks to the (negative) trend of Language in the United States and how powerful persons rather than encourage improved and better use of english, wield their influence to endorse the collapsing of standards across board. Is there a link between that and growing failure of thought and the narrowing of ideas in your country a la 1984?
The response from PG’s defenders eventually devolved to “If you want to do business with America, you must speak like Americans” and “Nigeria’s GDP is merely 1/70th of America’s, you are beneath our notice and we will not deign to answer you further”. Apparently, to speak like America/Americans is to scuttle our own linguistic abilities.
Now, do not get me wrong. America is my most beloved country in the world, and I hope to be granted an opportunity to live, learn, and work there in the near future, but there has been a deterioration in certain aspects of your national existence. Language is one of them.
I have recently been reading speeches of great American men of old (mostly speeches of a political nature) and have been paying attention to the dexterity in their use of the anglo-lingua. I have observed the way they wield the language so powerfully in furtherance of their goals. I am currently reading Paul Calhoun’s speeches in support of slavery. As I told my friends, the moral quality of his speeches are certainly abhorrent but their literary quality are unimpeachable. The speeches of Frederick Douglass, a self-taught escapee were also of a similar if not superior quality.
For a country with such wealth of history and examples to draw from, the current situation is quite saddening.
I know Paul more from his role as a computer programmer and advocate of an esoteric niche programming language. I also develop in a niche esoteric language from the punchcard era, but not Lisp.
This observation isn’t a criticism on your use of language, more an observation that informal use of English is under-represented in AI training and thus over-represented in AI output. At the same time, informal English is over represented in U.S. education and vey little formal, resulting in observable language use indicators of someone who may have used such a tool.
If any programmer looked at a piece of Lisp code Paul wrote last week and said “The language used here _indicates_ it was written before 1990” they would be semantically correct but factually incorrect. There surely are also other measurements that detect other indicators like British spellings that would drive the “likely AI” indicator downward, so don’t take Paul’s dismissal at face value.
Other respondents in the thread using appeal to ignorance are best ignored.
As a total aside, what language do you use?
It’s the language formerly known as Mumps.
Mixed feelings, because I get what both of you are saying.
Google Gemini recently came under fire because if you gave it a prompt like “please draw me a picture of a Viking”, it seemed like the program specifically avoided returning a picture of a white man. Example after example of horned-hat-wearing African men and Asian women abounded. At the outset: I thought that there was at least a possibility that it was a sampling issue – That because less than half of the internet is white or English speaking, coupled with how DEI has made a lot of stock photos unrepresentative, that we might have so colored the sample of the internet that Gemini relied on that those pictures were actually what the AI thought were representative of Vikings.
It turned out that Google had purposefully added lines that forced diversity on requests like that, and a quick edit of the code cleared it up almost immediately.
And in the current environment, maybe that was the most likely problem, but my original thought wasn’t facially absurd: AI relies on sampling, ChatGPT3 sampled about 600GB of text (About a million books worth). While that’s a lot, it obviously left a lot of holes and gaps, and the writing patterns are often noticeably rough. Those rough edges are being worked on, but they’re being worked on by expanding the sample, and training out wrong answers. There is a reason that chat GPT doesn’t talk in slang, ebonics, or spanglish… The samples being fed into AI is usually from published text. And published text is a little more superfluous than your average speaker or shitposter. It would not surprise me if AI got hung up on a set of words that it used much more commonly than the market at large, and those words listed don’t seem absurd examples.
So I think that it’s likely that Graham might have a point in a general sense… A piece written like that might actually be more likely to have been produced by AI that an average piece. Not to say that any specific piece, or all pieces including those word sets are AI generated, just that they sound like they might be.
Which I understand might come off poorly to a population that writes like that. It’s an interesting bit of culture that the dialect in a lot of African nations is very textbook proper and flowery, but it is very odd to a Western ear. You can complain about the slackening of American standards, and don’t get me wrong, they’re slack… But I’m not sure that your specific dialect was ever the norm… I don’t think the problem is degradation, so much as it’s mutation. Even here, you’ve written things like “Now, do not get me wrong” instead of “Now, don’t get me wrong”. The lack of a contraction sounds tinny to me, and reminiscent of a Star Trek character: Data – An android, an AI, incapable of using contractions. That show ran 40 years ago.
AI is going to be the social and ethical problem of this generation, and it won’t be close: All the noise and effort we’re spending on trans issues will pale against things like job replacement, automation, academic performance, and intellectual theft. This is a symptom of that: People are looking at ways to identify work produced by AI, and your local dialect is being impugned this time… But this is just a stop on the line.
Interesting observations, HT, as always.
jvb
It is unethical for XY to compete against double X chromosomes, and is in fact demonstrably physically dangerous to do so. The woke virus continues to spread because fear-based invertebrates are afraid of being called names.
Riley Gaines: I Stood Up for All Women after Losing to a Man | PragerU
I wonder if she would be so vocal if she had not been beaten by a biological male. I hope she would. The Williams sisters have confirmed that they could not compete against the leading male tennis players, not even close.
It is unfair for biological female to compete against biological males in events/areas where physical strength and/or prowess is a benefit *or it is not, regardless of the outcome. If Gaines had beaten Thomas, would that take away from her argument?
jvb
*Ed. Note: I am not referring to economic issues or other areas. This comment is confined to those areas where a biological male is going to be physically stronger than a biological female and said strength creates an unfair outcome. Specifically, I am referring to sports such as swimming, soccer/football, wrestling, baseball, golf, track and field, and volleyball, to name just a few. I am also talking about athletes at similar levels in their respective sports, which means elite male/female swimmers or runners or wrestlers.
I watched our son swim competitively and saw that, even at 12 or 13 years old, boys were physically stronger and faster than girls. For instance, in the 500 yard freestyle in one competition, the meet organizers had the idea that boys and girls would swim in the same heats. Boys finished sooner than the girls. Two years ago, our son swan the 200 meter butterfly in a regional competition, finishing first at 2:03.05 (setting a personal, team, event, and competition record), the next fastest boy/male was 2:08. Both were faster than the winning female swimmer in the same even.
I think the strongest condemnation of the status quo comes from the fact that she TIED Thomas, but Thomas was given the photo op and the award solo. Because he was trans, he got the credit for the women’s victory.
jvb,
First, according to J McEnroe, the Williams sisters would not do well against even triple digit ranked male players. Serena played an exhibition match in 1998 against the 203-ranked male player and lost. ”If I were to play Andy Murray, I would lose, 6-0, 6-0, in five to six minutes, maybe 10 minutes,” Williams told Letterman. “The men are a lot faster, they serve harder, they hit harder. … It’s a completely different game.”
Second, Thomas was taunting the women in the locker room by exposing himself. He is a world-class asshole but unfortunately he is the norm among trans-activists.
The problem with indulging trans-people is that the whole thing is subjective by nature and leads to, well, all the unnecessary garbage we see today, like women being seriously physically injured. Too many maladjusted guys are taking advantage of the woke virus thereby displaying a total disregard, possibly latent contempt, for the female athletes.
There are exceptions like chess where there can be male, female, and open tournaments.
The problem/controversy eventually goes away if the deciding factor is chromosomal. Taking the bottle away from the baby ends in a nap.
Taylor Silverman: The Trans Movement Is Erasing Women in Sports | PragerU