I’ve fired one of America’s most powerful lasers—here’s what a shot day looks like

The laser was used to study the physics of stellar interiors and fusion energy, among other things.

See full article...

 

[SomewhereAroundBarstow]

I know Ars has (rightly) been castigated for LLM usage in writing articles, but what you quoted is a normal-ass sentence for anyone who's graduated with an English degree or has written for a while.

But...but...ThE eMdAsH!!!!111!!!

/s hopefully obviously, but I bet money that's what OP thinks is the tell.

 

[SomewhereAroundBarstow]

Not germane to the story, but in that same 17-floor building if you visit one of the side stairwells heading down toward the basement you will find a very hidden, out-of-the-way restroom that is inexplicably accessible only through the stairwell. Used to be a top spot for me for peace, quiet, and relief.

If you are in the stall of that restroom and the writer of this article opens the door from the stairwell and asks "Is anyone in here?" DO NOT STAY QUIET.

 

[SomewhereAroundBarstow]

You'll never guess what ChatGPT is trained on, or as you put it, whose output it imitates.

 

[SomewhereAroundBarstow]

Various word processors automatically turn a normal dash into an em-dash.

I sure hope they don't. A hyphen and an emdash have different grammatical purposes.

A hyphen joins words or word fragments, while an emdash sets off additional information or indicates a pause. Some word processors will turn a double hyphen into an emdash, but I haven't relied on that for years as on MacOS an emdash can be easily typed with "option-shift-hyphen" and on iOS you hold down the hyphen until a selection menu pops up giving you access to a variety of different dashes.

 

[SomewhereAroundBarstow]

No I wouldn't guess...I know how it works.
As new material is added to its training material, changes in style caused by adaptation of human authors to AI will incresingly find their way into AI until, like the mythical poot bird, it will fly in ever decreasing circles and disappear up its own cloaca.

No writer who treats writing as a craft is going to imitate AI. If anything they'll go to greater pains to make sure they retain their own distinctive voice. What will happen is people who don't have any respect for the craft will increasingly outsource writing to AI and that output will then be ingested by AIs and thus AI output will become more and more homogeneous as the volume of AI-generated writing buries the human-generated writing. This will have the effect of making human writing stand out even more amidst the sea of dreck. Yet there will still be people saying "I can tell this is AI because it uses emdashes and the word 'quiet'."

To bring this back on topic, I am working on a grant proposal to put Sam Altman in a vacuum chamber and shoot him with a laser.

ETA: Nothing quite like writing about writing and using "their" instead of "there." Sheesh.

 

[SomewhereAroundBarstow]

MS Word will turn a space-hyphen-space into an em-dash, since it's super rare that you'd have spaces bracketing a hyphen that should remain a hyphen.

I've thankfully not touched Word in well over a decade and I'm gathering from these comments that particular one is a newer trick. Or maybe I never ran into it because if you are putting a space before or after a hyphen you are not using hyphens correctly. I was trained by a style guide to set emdashes off with spaces but I'm sometimes inconsistent about that in my own writing.

At that same job, for a while I had to take what I'd written in Word and mark it up as HTML. In Word. Thankfully I got a real text editor after a while because that was misery.

 

[SomewhereAroundBarstow]

It’s faster to just hit the hyphen twice—as I just did—which works on iOS, MacOS and reportedly Microsoft Word on all platforms as well.

I'm not sure about faster but I am sure -- having confirmed it just now in this reply -- that it doesn't work in Firefox on MacOS.

 

[SomewhereAroundBarstow]

Note that it’s a convention in British publishing to offset em-dashes with spaces, whereas in the US those are not normally used. Of course conventions are just that.

Yeah, we had an idiosyncratic style guide with some deliberately anachronistic elements to it.

 

[SomewhereAroundBarstow]

Because LLMs don't work that way. It would be like using an airplane for a submarine, simply because they are both roundish.

Of course there's a slightly greater chance of success in using an airplane for a submarine than there is in using a submarine for an airplane. Not great, but greater.

 

[SomewhereAroundBarstow]

MS Word used to drive me nuts with this, along with changing quotation marks into fancy open/close quotation marks. I had to fix people's documents for upload and our website of the time wasn't having all these fancy characters.

I worked for a certain company that had several years before I worked there tried and failed to expand into Japan. As part of that attempt they translated the entire website, with tens of thousands of products in the the catalog, into Japanese. That required reconfiguring the product database to use 16-bit character encodings.

While I was there, the need for more storage space for the product database came up. The database admins went to their boss and said "We don't need to buy more space. Because we don't have a Japanese version of the site anymore all we need to do is reconfigure the database to use the ASCII character set." (I'm sure they said "ASCII" but technically it was ISO 8859-1.)

The first anyone outside of the database team knew about it was when every curly quote, emdash, and curly apostrophe on the website turned into a question mark. That last was especially bad because it's used in the trade name of one of the biggest brands the company carried.

If I'd had my way those guys would have been out of work the next day. Instead they did a text dump of the entire database that I then spent the next night staying late at work running RegEx passes on to rewrite all of the permanently lost characters as entities ("—" and the like). The next morning they uploaded the text file to the database and "fixed" it. That was almost 20 years ago and I still get angry thinking about it.

Yeah, big lasers are cool.

 

[SomewhereAroundBarstow]

Wow. Switching to a non-Unicode encoding is a remarkably short-sighted decision, considering that the company might have wanted to expand into other international markets at some point. Hard to imagine that the database size savings from halving the encoding size of the text alone would have been significant enough to "solve" the storage problem in any era where a product website existed, especially if you also needed to store product documents and images! This must have been before UTF-8 was a practical option?

These guys had absolutely no business deciding all on their own what encoding to use but there was this really screwed up culture stemming from the failure of the Japanese expansion coinciding with the dotcom crash. IT leadership treated the online line of business as a problem to be contained even though (at the time) we were bringing in a quarter of the revenue and growing. It literally never occurred to them that as the person managing front end production I might have any opinions about, or even knowledge of the existence of, character encodings. But at that point I'd been working on web front ends for a decade and also had a stint writing XML DTDs. I'd learned quite a bit about encodings and their pitfalls.

But to clear things up a bit, I described it earlier as the product database but really I should have called it the online product information database. The real product database was primarily inventory control not just for online but for a hundred or so physical locations and had every product the company had ever sold since some point in the '80s in it.

The online product information database was text only and binaries were stored in a different database. Binaries might even have been flat file storage using product IDs as directories. It's been a long time and those kinds of nuts and bolts weren't very relevant to my job. We were delivering web pages in UTF-8 so there must have been a conversion happening in the publishing code. And I bet that's where the bad idea took root; one of these guys probably heard about that and assumed the UTF-16 was being converted to ISO 8859-1. It would have been just another example of how the online line of business had gotten too big for its britches during the boom and demanded a storage-wasting solution like that.

(I'm really not exaggerating about how IT thought of Online. My favorite example was was when the photo folks requested an external hard drive to take on a shoot and got back an email laying out in very twisted logic how if they got an external hard drive, IT providing support for it would cost the company a million dollars. I am not joking. A big part of my job turned out to be pushing culture change, and let me tell you, nothing burns you out more than doing that while also keeping production going.)

 

[SomewhereAroundBarstow]

With just one of these redundant but dramatic-sounding oppositions broken into into separate clauses, I'd be more on-the-fence, but with two, it's heavily likely there was some either some LLM involvement or the writer has himself been influenced by the LLM-driven trend of that form popping everywhere over the last couple of years.

FFS, you can't actually tell if something was written by LLM from simple stylistic choices, no matter how much you dislike them. If you go looking for AI-generated or -modified text you are guaranteed to find it everywhere whether it actually is or not.