Slop Cannot Replace Everybody (the Story of Perl and Universities)
Shown on the right-hand side is Mark A. Lemley, who blasted the deeply debt-saddled Facebook ('Meta') despite it being a client of his... it was publicly blasted for being a slop maximalist - rendering its practices indefensible (like Anthropic's)
Putting "hey hi" (or "AI") with "productivity" in the same sentence is participation in propaganda. Heck, calling image fusions (glorified CG) "hey hi" is also participation in propaganda. About 90% of the things today's industry (and corresponding docile media, working for or owned by the same industry) calls "hey hi" are anything but. Implying that slop is desirable is participation in propaganda.
Sobering Up
Get off the hype horse (or train). Some people made a career out of this horse. Some got Nobel recognition for it. Nobel prizes are not about science anymore, they're about money and power [1, 2, 3], so they get granted to those who serve the agenda of the wealthy and powerful.
Let's be honest with ourselves. As Linus Torvalds (an actual computer scientist, not a greedy poser from a slop vendor) put it some years ago, "hey hi" is 90% marketing hype. He was being too generous, he could say 99%. He said that before his boss, Jim Zemlin, took many more bribes from slop peddlers (for marketing campaigns) and Jim Zemlin's wife (and de facto boss according to his public talks), Sheela, worked for a slop company connected to Microsoft (she's still there by the way, she fled the fraud scene).
In this day and age we're surrounded by PR and snakeoil pushers, who also dominate the tech industry (even vaccines industry) despite never getting even a college degree. It is sickening, but this is what we have; just look what happened to the EPO's management. It's a parade of unqualified opportunists with "connections" and no principles. They try to replace examiners and even judges with slop. The effects are as horrifying as can be [1, 2].
Practically Speaking (and Thinking)
I honestly don't think Large Language Models (LLMs) can replace what I or my peers at Techrights do. We have a deep understanding of the issues we cover and we might as well nonchalantly interject that we actively avoid covering topics we don't fully understand. We ensure we get the facts right, we have reliable sources (including leakers and whistleblowers whose reliability we properly verify/vet), and enough common sense to determine what's true and what's bunk, without having to automate decisions related to that. LLMs cannot get the basic facts right. (No, I'm not married to someone born in 1868; it's gross, it's incest, it is necrophilia).
In academia, there's not much pressure to adopt slop (compared to GAFAM; many engineers there are forced to use slop for code and written text). But academia has a whole bunch of other issues, as we've covered before, as did Andy as recently as two months ago.
Having spent several years in academia (after my doctorate too), I have a rough idea of many if not most of the core problems, including the cartels associated with journals, conferences and unpaid peer-review sharecropping. The libraries are part of the problem (large publishing houses - which are more like the music recording conglomerates - plunder tuition fees) and my associates in Techrights can explain this a lot better than me, having seen it firsthand in academia.
At the moment there are some paid shills of Microsoft - pretending to be "academia" - pushing/peddling slop for scholars and even for judges. Oh, dear. This won't end well, will it?
Attack on Technical Communities
Similarly, we've noticed some people who intentionally or unintentionally cause fatigue in Perl (which we use heavily in running the site and Gemini capsule).
Killing the Perl community with slop (in "agents" clothing, a fancy name for bots) is the last straw for some contributors or mere "consumers" of CPAN*. They are, to borrow a term of MAGAts, flooding the zone with crap to ruin the actual science with fake chaff and bot activity/ies. The same is happening in academia (as a peer-reviewer I'd hate to look at slop submissions, just like teachers loathe marking up exams in which students/pupils cheat with LLMs), but there is thankfully sufficient resistance ("The [slop] scientist: now academic papers can be fully automated, what does this mean for the future of research?").
Now, going back to Perl, suppose it is now besieged by slop fanatics and boosters of Microsoft GitHub (plagiarism by slop), who are addicts or users of proprietary software. What can be done about it? "That's one of the many brutal blows meant to destroy the community and thus Perl," an associate of ours opines. Perl should definitely focus on people (the core people, productive contributors, not "yokes" to peers) and quality, not quantity of sloppy slop-slop**. Perl does not need more code, it needs more people and better maintenance***. Adding slop would lead to burnout and less interest in genuine participation.
More Can be Undesirable
Likewise, academics rely on their reputation to attract interest and grants. It's not about how much text they type (or publish), it's about overall impact and published (after real, extensively peer-reviewed) work. Writing more and more words isn't the way to get ahead in a world with LLMs and all sorts of online bots. Distinguishing oneself by quantity is misguided. Regarding slop, in general (also in blogs): writing a lot of text is not hard, many people can write and write and write all day long (given time and search engines etc. for 'research'), but earning and keeping an audience relies on some proven track record and this is what really counts. A million words read by zero people is still 0, as 1,000,000 x 0 = 0.
"See the recent post in Daily Links about the actual purpose of scientific papers," an associate has said. "It's not about the production of text."
"The text is not the product," Crooked Timber explains.
To quote from the post:
There may be situations in which texts are really products in and of themselves. I wanted to provide examples (certain types of cheap fiction writing? user manuals? the small print in contracts?), but the longer I think about it, the harder I find it to come up with examples that would really fit. We treat texts as products; they get bought and sold (think of everything around copy right and IP). But in reality, texts are almost always something else. Here is an incomplete list of what texts can be: [...]
Source: Crooked Timber
Our de facto policy is, not all articles need be long, nor all videos need be movies; the point is getting key points across, maybe link to what was already covered before; so people can cruise by what matters and skip the cruft; when boycottnovell.com started in 2006 the blog posts were not long (it was new, hence a much smaller audience), but the points resonated with people and many decided to stick around and help out.
In academia, certain professors can fill large auditoriums and convince thousands of peers to read their papers; in the area of copyright and patents, Prof. Lemley comes to mind. He and other peers in that area contacted me in the past and we've generally mentioned him a great deal over the years. We do have respect for Prof. Lemley, but we didn't appreciate it when he sent paid PR agents our way. He was trying to outsource of not automate his "marketing".
Signal First
Papers and abstracts in academia (online) don't like to add view counters - or not anymore anyway - not because they're impracticable or due to bots but because they hide how few humans bother to even open the abstract page (for 90%+ of all papers) as it would discourage further engagement/participation. Not all who open the page read the abstract. Those who open a corresponding PDF file might, at best, stare at some figures. And the overall totals? Usually dozens, depending on the discipline and platform. The same goes for conference talks that events upload to some social control media like YouTube. Over 90% of the talks will barely be accessed by anybody (unless it's some very high-profile speaker/s).
The solution to this isn't more words, with or without slop (LLMs). The solution is better signal-to-noise ratio. Concision can help.
LLMs are basically SCIgen on steroids (or "autocomplete on steroids"). Have we learned no lesson(s) in the past 20 years (since SCIgen)?
When it comes to publishing, it is not the volume but the substance, as bytes of text are not scarce.
Back to Basics (Reality Check)
People who say they publish books made with LLMs won't disclose the fact that nobody reads them. They just want to say they "published a book". The same goes for slopcasts or "podslop" (some use the latter term). Making LLMs spew out some text and then using voice synthesis to read the text will never exceed the value of a real person expressing real ideas using a real voice. But they try to trick platforms with SEO and sheer quantity.
The same goes for coding. This brings us back to Perl. What Perl needs is more people, not more code. What academia needs is better minds, not more bots.
Quantity where abundance exists (and can prevail) is without merit; quality is what people opt for as they have limited time and patience. █
____* Apropos Perl under siege, it was the original "P" in "LAMP" for many years before the Microsoft whisper campaign steered people to PHP. Also note the process to even attempt to donate to the Foundation. Though that seems to have improved, slightly, of late so perhaps there are people on the inside who are aware and actually working to save the project. From workflow perspective, it is now actually possible to donate. There are many problems with Stripe, and really the other payment services, but at least they function somewhat smoothly and reliably unlike PayPal.
** The number of lines of code or SLoC has long since been debunked as a productivity metric further, in regards to both security and maintenance costs, as each line that exists is a liability, so shorter is better.
*** More emphasis on or a better understanding of the problems cause by toeing the Microsoft line by deepening ties to Microsoft GitHub is needed. Perl needs more people but burrowing deeper and deeper into Microsoft lock-in alienates the very people most likely to use and support Perl. The same for trashing an ever-increasing number of their essential Web pages and web services in gratuitous JavaScript.