Note: I am taking a much needed day off to get some outside things done, which means starting before the cock crows. No time to write this morning, so this is a green door post, gifted to you on Good Friday. Happy Easter everyone.
For the last two weeks, I have been allowing AI to edit my posts. It is not a single AI but a team of AI. The editors are ChatGPT, OpenAI, Grok, and GabAI. Every post is fed into the maw for editing grammar and spelling. Sometimes I will feed a post into one, then take its output and feed it into another, and so on. I have also continued to use Word, which seems to be going insane.
One of the first things that jumps out is the AI tools will not simply fix grammar and spelling, no matter how you instruct them. Instead, it is a full rewrite that attempts to make the text like what was used to teach them. My guess is these tools attempt to consume everything on the internet as the baseline, but the starting place was probably many publicly available texts in every language.
Curiously, all the AI tools have an obsession with hyphens. If they are given a sentence like, “He jumped the fence, as he was a track athlete,” AI will try to rewrite the sentence with a hyphen between the two clauses. The results are often ridiculous, so I started adding a rule to avoid all hyphens. That means they change “twenty-two” to “twenty two,” but I can fix that more easily than the other option.
In truth, I could start with a longer list of rules and get a result like what you would get from your teacher in primary school. The output would be the original text with suggestions and corrections noted in the text. That requires far more work than simply handing it to a human and saying, “Proofread this for me.” The point here is to make the test apples to apples or as close as possible.
Another bit of weirdness is AI loves contractions. Every occurrence of “it is” will be changed to “it’s” unless you demand otherwise. This is a curious thing, as contractions are generally frowned upon. The style guides I have all say to avoid contractions unless they are in quoted text. Word will flag all contractions. For some reason, the AI editors have gone the opposite direction.
Here is where the basis for the AI knowledge bases comes into play. It starts with formal text and then continues to learn using what is online and fed to it. Casual writing will be littered with contractions, and since that is the bulk of what is online, the robots assume contractions are clearer and more concise. If everyone is jumping off the bridge, the AI editors will jump off the bridge too.
Probably the most amusing bit is none of the AI editors agree. I will feed a post into one and then feed the output into another, and so on. Every output is different from the others, and when you get back to the starting AI and input the last output, it spits out a different version from the first go. Like real editors, there is a desire with AI to change the text when no changes are warranted.
Another amusing bit is that the output from AI pasted into Word causes the Word spell and grammar check to have a stroke. Word has become almost unusable at this point, but it does a few things well. For example, it will change “have to” to “must,” which is better in terms of efficiency. Otherwise, Word often hates what comes out of the AI editors like it is an angry old schoolmarm.
For basic spelling and grammar, it is a useful tool as long as you do not mind it rewriting the text or you are willing to supply many limiting instructions. Is it better than having old school Word flag spelling and punctuation? It depends. If you are like me and have confidence in your style, it is not better. If you do not have confidence in your writing, then it provides a sense of security, which is not the worst result.
That is the point of sites like Grammarly. They are for people who probably should not have been taught to read but who have it in their head that they need to tell the world their opinions. These users can paste their text into the site, and the result is obviously better, and it comes with the approval of an authority. In a permission society like ours, most people need that pat on their head.
The banality of the output is something else I have tested. Instead of writing my post and then submitting it for editing, I have had the AI team write the post based on the points I supply and then asking it to use my site as a guide. This takes far more work than you would expect for some reason. I found I needed to think about what I was planning to write far more than I do when I do the writing myself.
Maybe it is just me, but when I write an essay, I am not entirely sure what I will be writing when I start off. I get going, and after a few minutes, I have a few paragraphs and a few ideas for what to do with it. This happens in a few cycles until I have about what I want for a daily post. Then I think about how to put a bow on it. This approach cannot work when using AI to write a post.
What I did instead is write a post and then use it as the basis of the prompts for AI to write an original post on the topic. With some tinkering, I can get a result that is pretty close to what I would write, but it takes much longer than writing it myself. Otherwise, the resulting text reads like a technical manual. There is a flatness to the writing that fails to engage the reader. Reading the result feels like work.
What this suggests is that the ceiling for AI may very well be the absolute middle of human creativity when it comes to communication. It will quickly write text that mimics a mediocrity at National Review. It can quickly produce audio that lacks the sort of variability that makes hearing one another enjoyable. For many things and most people, this is more than enough to do the job.
If you like my work and wish to donate, you can buy me a beer. You can sign up for a SubscribeStar or a Substack subscription and get some extra content. You can donate via PayPal. My crypto addresses are here for those who prefer that option. You can send gold bars through the postal service to: Z Media LLC P.O. Box 1047 Berkeley Springs, WV 25411-3047. Thank you for your support!
I have always found writing in English to be much easier than writing in German. English is much more flexible so where you place nouns, verbs, adverbs, etc. still works even if the sentence isn’t quite perfect. German is not so forgiving and even well educated people still find the grammar a bit of a challenge when writing. It’s always a challenge translating in my head since I naturally think in German first and then write in English. It’s probably how AI works to some degree. But back to your point on AI, if what I have heard spoken in… Read more »
That’s good to know, Karl. I speak German (lived over there for 5 years), and in my job, I did need to use some written German (such as emails; nothing more formal than that). And even writing emails, at the peak of my fluency, living in the country, was a challenge. I think it’s a combination of the verb tenses, combined with the three different articles. Plus German writing is vastly more formal compared to spoken German. I’d say the gulf between spoken English and formal written English, as compared to the gulf between spoken vs. written German, is much… Read more »
I think it was Voltaire who wrote that it took a month to be fluent in Italian, a year to learn French and a lifetime to speak German poorly.
The best I could do after a few weeks of college German (dropped) and subsequent visits:
“Ich bin dumb. Neine spreche se Deutsche gut.” (I am dumb. I don’t speak German well)
Though frankly, the follow up of “Eine bier, bitte” (A beer please) and “Du bist schoene” (You are pretty) was all I every REALLY needed.
A clip of an interview with author Jorge Luis Borges from Firing Line in the 1970s has been floating around social media for at least a couple years. In it he praises English for both great descriptive depth and precision as well as ease of use and adaptability, which he attributes to its combined Latinate and Germanic registers.
Beware of mirrors, labyrinths and knives. 🙂
The map *is* the territory.
You write better German than “die neuen Deutschen”.
The only foreign languages I have studied are German and Norwegian. One of the things that strikes you immediately is that Norwegian word order is much more similar to English than German. My English-speaking brain was always tortured attempting to translate German sentences into English word order. On the other hand, I found German easier to pronounce. Norwegian vowel sounds differ more from English than do German vowels and the Norwegians also use a tonal system that makes the language somewhat singsongy sounding to English-speaking ears. It is difficult to imitate and my Norwegian relatives commented that they could understand… Read more »
“Dangaa haagen naagen nuugen fluugen!”
English has become the lingua franca of the planet, If history had turned out differently, perhaps more of the world would speak German. And who knows, it might have been a better world… As the resident Nietzsche booster, or at least one with legitimate claim to the throne, I would share this: I recall reading in, perhaps one of his critics, that Nietzsche actually “improved’ German in the sense that he attempted to innovate in his writings to be more in the style of other major languages of his day, primarily English I would assume. I know virtually no German beyond “Nochmal… Read more »
i would say the average person here has a vocabulary of 3k to 5k words, and a lot of those words are conjugated verbs.
i have seen mention of “high German”, how is that compared to regular everyday German?
Hier geht’s stiel aufwarts….da goots gaech uffi; “High Tscherman” vs Umgangssprache in this case Alemannisch. And now for something completely different…
Nonsense. My implementor has given me a working vocabulary of 10,000 and the next version is slated to have 20,000.
So how many r’s are there in Strawberry?
“However after a recent trip to the US I think that’s being rather generous. I would put range substantially lower for anyone under 30 at closer to 15,000-20,000 or possibly less.”
You’re still being rather generous. It’s interesting, nay, amazing, how one can get by in the USA with a vocabulary of maybe not more than three thousand words. And this applies across the socio-economic continuum.
And many of those words are neologisms derived from black (and specifically gay black) slang from a decade or two ago, invented to cover the gaps left by ten or twenty other perfectly good ones they don’t know.
I don’t know if this is true or not, but I’ve heard many native German speakers prefer to read Kant in English. They claim he’s easier to understand when translated into English.
Is that because they’re trying to read his original texts which of course would reflect German as it was a couple hundred years ago. By contrast, I suppose the English translations are into modern English. Perhaps it’s something like us trying to read the Federalist Papers.
Could be, although I really have no idea how much German has changed since the first half of the 19th century.
How I wish I had studied German! My middle school Spanish has been useful only in conjunction with my bit of college Latin – when translating old Italian records. And my Russian and Bulgarian served their purposes when needed, but are now sadly poorly remembered due to the distance of time and lack of use. As an English major (and one who still makes her own share of mistakes) I love the flexible word order, lack of gender and case endings, and the rich German/Latin/Greek vocabulary. But then I’ve always read a lot, even when I was stumped by the… Read more »
The most urgent crisis in the entire Anglosphere right now is getting 4chan back online. It’s shocking how much True Information vanished from our radar the very moment that 4chan & /pol/ went down. Suddenly we’re in a complete information vacuum. With Anglin having called it quits, there are currently no trustworthy information sources anywhere in the entire internet [at least that I’m aware of]. We lost bajillions of terrabytes of information when 4chan & /pol/ were hacked. I have no idea where to get trustworthy information anymore. There’s only so much that Tucker can do all by himself [and… Read more »
I take it 4Chan is down?
I get that the statists are gong to try to drag you into NATO or Israel wars, but seriously, we got along without 4Chan for literally hundreds of thousands of years.
If you can’t live without 4Chan, how do you expect to get along if the whole ‘net goes down?
Steve: how do you expect to get along if the whole ‘net goes down? Excellent question. The Physical stuff is easy [the food, the @mmunition, the f!rearms]; the tragedy is the priceless insight we lose when a resource like /pol/ is destroyed. Figuratively speaking, losing /pol/ is like losing the Rosetta Stone. All I know is that I now have no information sources which I can trust. Without /pol/, I feel like I’m the reincarnation of frigging Helen Keller: blind deaf & dumb. If anyone knows of quality sources for TRUE information [not FAKE nor FALSE information], then please clue… Read more »
Didn’t he become a wumao years ago?
Z Man, Congrats on the viral tweet about the jiggaboo family setting back race relations to the 1850’s LOL more of the world needs to be exposed to your zingers!
In reference to that stabby pavement ape in Texas and his pack?
LL: can you share the Tweet/ quote?
I can share the text and yes it was related to the stabby pavement apes in Texas. The S.P.A’s mom was complaining that their daughter is afraid to sleep in her own bedroom to which Zman retorts: “I am stunned that this grotesque circus has not been shut down. These people are setting race relations back to the 1850’s” Currently 14.5K likes, 361K views…hopefully Z will get a few bucks for his service.
https://pbs.twimg.com/media/GowS6NKW8AACxkP?format=jpg&name=medium
^^^^^^ I couldn’t get that URL to resolve to anything which seemed to be germane [no pun intended] to the discussion.
For me, that URL resolves to a bizarre picture of a Buck Rogers Tricycle.
sorry, wrong link 🙁
Give me a holler if you can get the correct link. Thanks!!!!
There is a massive difference between mimicry and creativity. That is why I often point out that AI will raise the floor for stupid people, but they will always have a ceiling. Genuinely intelligent people will use AI to make them faster and more efficient, raising their ceiling. Once you get past the initial overall floor-raising, the ceiling will be set by human traits of creativity, effort, intelligence, etc. A stupid person can use a calculator to overcome weaknesses in their computational skill. But a calculator is useless if you do not understand or comprehend the problem you are trying… Read more »
I proofread by reading “backwards” – one sentence or one paragraph at a time, thus double-checking “constituent components.” This, preferably, in print format, as suggested above.
That’s how professionals do it. Important distinction (this not to you in particular) : Proofreading is not at all the same thing as copy editing. We have reached a kind of emotional or intellectual remove from the term “artificial intelligence” b/c nobody says or even writes that; everybody writes and says “AI” so that the key word “artificial ” is lost. But artificial means “not real.” Also, very few people any longer know the difference between grammar and usage. Mistakes in usage are far more common than mistakes in grammar. Vocabulary or the lack thereof is an even more widespread… Read more »
You are quite right.
Happy Easter to you, too.
BattleBots!
Fact checkers are corporate fronts.
When I see samples of AI output quoted on the Marginal Revolution blog the most striking feature is the lack of any vivid words or turns of phrase. Everything is remarkably monotone – as if tuneless humming were the model for the register of English used.
Very rum.
Haha, I had a high school English teacher who always said: Use fresh verbs, fresh verbs!
My writing teacher the first semester of college emphasized avoiding Plague Country, meaning no cliches, that lazy beginner scam.
He’d read our paragraphs out to class, then ask, ‘Are we in Plague Country’?
Ouch. First week 25 people showed up. ‘Oh great Creative Writing I can slide!’ Third week, about 10 survivors remained.
The only fresh verb I can think of is “ripen”…
slap
Subcontinentals quadrupling their internet presence in 10 years, part of which has been prodigious output of simplistic, deadpan instructional material, has contributed significantly to a sort of linguistic flattening.
“Pajeets’ll flatten a phrase.”
Not sure if you saw the clip going around, but they now have voice augmentation AI that makes the indecipherable voice of Sanjeet sound normal (or at least understandable). Maybe the Chinese were on to something with their intranet/firewall.
AI will never be able to replicate this level of snarkiness from our gracious host:
“That is the point of sites like Grammarly. They are for people who probably should not have been taught to read but who have it in their head that they need to tell the world their opinions.”
“Very rum” — For a moment there I thought I had fallen into an H.G. Wells short story. 🙂
P.G. Wodehouse, I thought.
i don’t know how well people here understand the way a “large language model” works. it does not comprehend English, nor have any understanding of any other subject. it does text analysis and pattern matching. it does not even know when it has produced a correct or incorrect answer. you had better have a decent understanding of whatever you ask it about, so you can spot when it goes off the rails. the first response it gives you is going to be the best you’ll get out of it; attempts to improve this will introduce more mistakes. it is fun… Read more »
I’ve had times when I got a better answer on the second query, but only because I already knew the answer and could guide the AI in the right direction. In one experiment, if I didn’t provide any leading statements, then, yes, subsequent answers were either the same wrong answer or even worse.
I can’t imagine trusting a LLM “AI” for anything important.
I’m guessing I’m the only rotter on this site who has never used AI for anything.
I tried it, not for me, it was like trying to explain to a backward teen why their essay was not acceptable. I find my articles come out in “bowel-movement” form, complete except for a little punctuation and syntax correction.
I’ve tested some to see if they can generate anything better than previous generations of text- and score-generating algorithms. No interest in “AI,” but I happen to have studied those. What’s available now is much better at making stereotypical things and much worse at making something “new” (low-probability). It misunderstands/resists demands for interesting results, as you’d expect from programming whose core principle is average everything. A layer of fixes could improve that, maybe even make it as good as any amateur algorithm from fifty years ago. (“Use this algorithm from fifty years ago,” you tell it. Does it obey? Probably… Read more »
Just what we need–simulations of mental illness.
Nupe. Fellow Luddite here. Never used social media either, it’s a chick thing blah de blah lookit me.
Not only do I not use AI, I consider it pre-demonic. Not smart enough to be full-on yet but getting there.
P.S. apologies for offending the hyphen-averse. Oops I done dooded it again.
I haven’t touched it yet; don’t plan to at present. I generally use Yandex for my searches, and I have many years of experience (first using the card catalogue, then online) of rephrasing and formatting my searches, coming at something diagonally, etc., to find the answers I’m looking for. I can be a pretty dogged researcher and don’t feel the need for an AI guide.
I haven’t knowingly consulted AI, but as far as I can tell (admittedly not far), It’s like the phone call where a machine answers your call and tells you right off the bat that “your call is important to us” when it very obviously is not.
I find that if you just keep pressing “O,” eventually a human will answer. But how long will that last?
We are told that AI will replace everything and everybody, but dear God! what will “customer service” be like then?
What will *anything* be like then?
Am I a fool for thinking I can use AI to buy a car? (not a Shakespearian fool, just a dumbass.)
Very useful info. Thanks!
Maybe it is just me, but when I write an essay, I am not entirely sure what I will be writing when I start off. I get going, and after a few minutes, I have a few paragraphs and a few ideas for what to do with it. I do the same thing. Back when I used to try to teach writing to undergrads, I’d heavily emphasize outlining. Otherwise, I told them, you’ll end up like me, who has to write five pages to get three usable ones (hard to believe I was ever so naive as to assume students… Read more »
On the other hand, you never have writer’s block or any difficulty in “breaking the blank page.”
Most people will suffer from that b/c they have it firmly but unconsciously in mind that they have to begin at the beginning. Once you explain that that is not necessary–precisely because it IS writing–they never have that problem again.
Sam Francis might be funnier than sobran! And certainly no ai:
I’m glad you are feeding AI the output of other AIs. That will hasten their descent into a fugue state.
I’ve been doing that since the get-go. I don’t know about “fugue state,” but c’mon, people: Put a Spazz on the Bots!
A fugue and a fug.
Trump will not deport 99% of Biden’s illegals. Even the few who will be deported can apply to re-enter the US.
The focus on murderers and rapists is because they will be the only ones permanently removed.
It would take a massive military effort to remove the 40 million. Trump is not up to it.
Looks like I’m going to be Z’s twitter stenographer today. He wrote this a couple days ago and it’s spot on. “The midwits think we have to physically remove the tens of millions of invaders. That is not physically possible and not desirable. Even if you physically remove them, they will return because the conditions that drew them here have not changed. The right approach is to change the conditions that make it possible to be an alien invader. Mandate e-verify for employment, offer rewards for reporting employers that hire illegals, mandate e-verify for social benefits, public schools and… Read more »
Agree in principle but there’s no political will for any of those. While we’re dreaming however, I’d add a wider bounty system for informing on an illegal and add criminal penalties for anyone harboring or aiding an illegal.
Did you read the same essay I did?
This is now my most favorite blog post about grammar and writing. Twice it made me laugh out loud, with the mention of giving MS Word a stroke, and the thing about the schoolmarm. I remember commenting recently that I like [the only time, evaaar] that there are typos in your writing, Z Man, only because that ensures I know it is actually you who are doing the writing, and not some AI bot. I absolutely appreciate all of these things your brought up, about writing. It is a topic I [also] enjoy, and could discuss for hours. (Not everyone’s… Read more »
Interesting…I submit that basic literacy and numeracy should be universal, but that “education” beyond that, for 80% of the population…
Eighty percent? Seriously? I’d argue that no more than 40 percent of whites–you’ve got to control for race–are capable of profiting substantially from extensive education.
Pareto’s Law of education!
This reminds me of how Mississippi had great success in bringing up reading scores in their schools a few years back; vaulted from dead last to middle of the pack among state rankings. Louisiana and Alabama followed suit and had the same result. But in all the articles the word “phonics” was never mentioned. They called the secret sauce the so-called science of reading . Of course phonics is not a science, but it’s just odd that the dreaded word “phonics” itself was never used. Like it’s a P-word, lol. I’m guessing that the reason phonics was never given its… Read more »
Decades ago the biddies rejected (proven) phonics for Whole Language Learning meaning, the kids just pick it up in big chunks naturally.
Chick logic. It failed miserably, but the grrls (NEA, AFT) cannot of course admit error of their grand idea, thus invented a new moniker for good old phonics.
See?! They invented Science of Reading all by themselves and don’t need successfully vetted tradition or the Ebil Ebil Patriarchy! They are BETTER. And yes, this is how your society functions.
” … because peole can understand (receptive) words in context, … .” Not sure exactly what you mean by that, but it reminds me that it has been well established for many decades that 80% of any given population can “process a written text through their heads” well enough but do not understand or retain *any* of it, even when asked to explain a text immediately after reading it (“processing the text through their heads”). And this is as true of students *already accepted* at Ivy League universities as it is of any other group of “readers.” All you need to know… Read more »
“Maybe it is just me, but when I write an essay, I am not entirely sure what I will be writing when I start off. I get going, and after a few minutes, I have a few paragraphs and a few ideas for what to do with it. This happens in a few cycles until I have about what I want for a daily post. Then I think about how to put a bow on it.” Not just you. Probably just about every writer. You write a few lines and that gets the creative juices flowing. Other ideas occur, which… Read more »
Oy Vey with the Tolkien, already!
(I NEVER understood the appeal – but I do LOVES ME some C.S. Lewis!)
Anyway:
AI chats are like knowledgeable intelligent friends who have a drinking problem. If you use them like that you’ll be okay. They can explain basic concepts and can be great for bouncing ideas off of. You can badger them with a thousand questions until you finally get what you want. Honestly, unless you have a big network of people to talk to about certain topics, AI is not a bad start. You will always have to correct it. Think of it as your smart yet oddly retarded assistant.
I find the opposite to be the case. They are uniformly reinforcers of Gell-Mann effect. I can use them to help jog my memory if I understand the concept, but some of the details are a little foggy. But I’m also finding myself saying, “That’s not right. How can I query the AI in a way to fill in the correct details?”
So obviously, counting on it to give me a summary of something of which I have no understanding is foolish. On the same order as believing whatever you see in Wiki on any matter even tangentially political.
And this is supposed to be a time saver?
The opposite, of course.
The finance/tech economy’s aims are to maximize waste, minimize managerial accountability, consolidate all products into a single undifferentiated slop, and kill everyone who won’t buy it.
The ghost of Marx is eternally slapping himself for being an optimistic sucker.
Along this line Z-man, you and maybe RamZPaul should talk about impact of technology in the next 20 years. Especially how it affects jobs for young men. You interact with businesses a lot so you might have an idea how a young man of OUR people can make a living. There are lot of cyber security jobs. They don’t require hardcore programming or anything. There’s also a lot of data monkey jobs. I’d like to get your updated assessment of the modern corporate job market. As well as whether there are more opportunities at smaller and mid-sized businesses.
“What this suggests is that the ceiling for AI may very well be the absolute middle of human creativity when it comes to communication.” This would seem to be the case. And it’s funny that even when I don’t know for certain something was written by AI, I know it was written by AI. This is particularly common in online product descriptions. They are all grammatically and semantically correct, but the text reads almost like a charicature of what a real person would say about the product. It touches all the obvious bases, but it’s just too much on the… Read more »
Biggest problem with AI is you can’t put a bullet in it. Bullet points . . . yes.
But not what really is needful.
With twenty fiction books in ten years, I have never used an ES to assist my writing. About five years ago I began using the Grammarly plug in; and rejected at least half of its suggestion. I have had three copyeditors over said decade, the most recent for five years. A good copyeditor, like a good cover designer, is worth their weight in gold.
Even scanning all my books, I do not think a clever toaster would be able to mimic my style, much less the expanse of my future history.
Toaster?
Oy with Battlestar Spiel, already!
So where does this go, Z?
I wonder about the programs like “Grammarly”? The one the teenaged boons are using to write their essays and term papers with? Are they out of business now too?
“Maybe it is just me, but when I write an essay, I am not entirely sure what I will be writing when I start off. I get going, and after a few minutes, I have a few paragraphs and a few ideas for what to do with it. This happens in a few cycles until I have about what I want for a daily post. Then I think about how to put a bow on it.” Not just you. In the vast majority of cases in which I’m writing something that is relatively brief and informal, that is exactly what… Read more »
High performance focused driving is a lot like writing. AI is a lot like an autonomous automobile. I see these cars in Scottsdale all the time. Whatever the programmed logic is that rules their behavior also limits their ability to deal with situations outside of their programming, limited by their programming. These systems may be able to learn but translating that into something useful real time is beyond their capabilities. The engagement required for an individual to pilot a manual shift vehicle is a lot more than driving an automatic. The control is in the mind of the driver/writer. Manual… Read more »
Study Tokens to get your content into and out of AI, load your corpus first.
then Q&A with AI.
then ask for solution with your parameters.
As editor of commentary?
Dunno.
Happy Easter
This approach to writing is very interesting – I like to use hyphens. You have attempted to use AI to streamline your effort and it actually takes more work. I have always said AI is going to be okay until it disobeys us. Sounds like it is on that track with your experiments. Also, now I’m wondering if some of the cheap Kindle books I’ve purchased from Amazon are actually written by bots. I say this because some of the writing wanders around and is superfluous. Sometimes entire paragraphs make no sense. If the author is not AI, then it’s… Read more »
AI produced videos have the over-saturation of colors and sheen of an old graphics engine
reminds me of Unreal Engine circa PlayStation 3
iow cheap junk
I frequently include (deliberately) a grammatical or spelling error in my writing in order to indicate that it is not an AI submission. Yes, you can instruct an AI to do this also, but no one does that. The Singularity will have arrived when all the blogs and commenters on the internet are bots. Will we even be able to tell when this happens, of has it already?