The younger generation loves short-form, to-the-point stuff. Which is the exact opposite of what the current crop of GenAI makes. In a tiktok video, every sentence, every word counts because there's a time limit. If people don't engage with your content in the first 3 seconds, it's worthless. The video linked in another post starts off with 15 seconds of complete fluff. You'd have better engagement if you have a guy opening with BIIIG NEWS!! LABOR DAY BREAKFAST GOES HAYWIRE!!! and hook people.
GenAI is great at generating "stuff", but what makes good content isn't the quantity. What makes good content is when there's nothing left to take away.
What's funny is every GenAI "incredible email/essay" would be better communicated with the prompt used to generate it.
I’m only half joking when I’ve described ChatGPT-authored emails as a uniquely inefficient transport format.
Author feeds bullet points into ChatGPT which burns CPU cycles producing paragraphs of fluff. Recipient feeds paragraphs of fluff into ChatGPT and asks it to summarise into bullet points.
> Author feeds bullet points into ChatGPT which burns CPU cycles producing paragraphs of fluff.
GOOG, AMZN, and MSFT reportedly need to use nuclear energy to power the LLM farms that we are told we must have.
One must ask who (or what) in this feedback loop of inanity is doing the most hallucinating.
[1] _ https://apnews.com/article/climate-data-centers-amazon-googl...
[2] _ https://www.npr.org/2024/09/20/nx-s1-5120581/three-mile-isla...
Then users turn around and feed the fluff into energy hungry summarizers because who has time for a 5 paragraph email that could’ve been a three point bulleted list?
It would be a net win if it could normalize sending prompts instead of normal communication, which is not far in terms of useless waste of energy and space to LLM output that emulates it.
Similarily, MSFT recently announced the upcoming ability to clone your voice for Teams meetings. Extrapolating, in a few months, there will be Teams meetings which are only frequented by avatars. At the end of the meeting, you get an email with the essential content. Weird times ahead.
I think you are on to something there. I've heard executives talk about their current AI flow and it all sounds like summarization.
There's an increasing amount of prose written that will only ever be read by LLMs.
"Your essay must be at least X words" has always been an impediment to truly good writing skills, but now it's just worthless.
The way I explained it when I taught English 101 to first-year university students: any substantive question can generate an answer of a paragraph or a life's work; in this assignment I expect you to go into this much depth. Of course, good expository writing is as to-the-point as possible, so the first hurdle for most students was eliminating the juvenile trick of padding out their prose with waffle to meet an arbitrary word-count. Giving a word-count to an AI seems (currently) to activate the same behavior. I've not yet seen an AI text that's better writing than a college freshman could be expected to produce.
> Of course, good expository writing is as to-the-point as possible, so the first hurdle for most students was eliminating the juvenile trick of padding out their prose with waffle to meet an arbitrary word-count.
This is the most beautiful sentence I’ve read today.
Problem is, I doubt many people overcome this hurdle - that "juvenile trick" is pretty much the defining quality of articles and books we consume.
Indeed. We need better English teachers. :-)
We won't get them unless we appreciate both teaching and the Humanities more than we do. I was good, but by no means the best (75th percentile, maybe?). I loved doing it, but changed careers to IT because I'd never have been able to support a family on what I was paid.
A culture which pays teachers poorly, treats them with disrespect ("those who can do..."), and relentlessly emphasizes STEM, STEM, STEM is one that's slowly killing itself, no matter how much shiny tech at inflated valuations it turns out.
I don't know how it is elsewhere, but where I grew up we had minimum word limits on pretty much all essays. Doesn't matter if you can say what you want to say in 6 sentences, you need 4000 words or 2 pages or whatever metric they use
Oh, of course. Length requirements are important, for the reasons I explained up-thread. However, if teachers accept any old thing ("padding") to reach the count, then that metric is arbitrary, which (justifiably!) makes students cynical.
If a student can say all that they want to say in six sentences then they need to learn more about the topic, and / or think through their opinions more deeply. Teachers who do not take that next step are bad teachers, because they are neither modeling nor encouraging critical thinking.
In some places the majority of teachers are themselves incapable of critical thinking, because those who are leave the profession (or the locale) for the reasons in my comment above.
[Edit to add]: Please note that I say bad teachers, not bad people. Same goes for students / citizens, as well. The ability to think critically is not a determinate of moral worth, and in some ways and some cases might be anti-correlated.
Wish my high school English teachers had taught that. I remember fluffing essays to get to a minimum
College admissions essays on the other hand had the opposite problem - answering a big question in 500 words. Each sentence needed to pack a punch.
Don't get me started on college admissions essays. Rich kids pay other people to write them. Poor kids don't understand the class-markers they're expected to include. If AI consigns them to the dustbin of history it might be the first unalloyed good that tech ever does.
Clarification: "that tech" meaning the direct antecedent: "[AI] tech", not tech in general. I'm a Humanities guy, not an idiot. :-)
One of my favorite Mark Twain quotes comes from one of his correspondences: 'My apologies for such a long letter, I hadn't the time to write a short one.'
Probably not by Twain, but still a good quote. https://quoteinvestigator.com/2012/04/28/shorter-letter/
I never had that requirement outside the first years of school- where it’s more about writing practice than writing actual essays. After it was always “must be below X pages”
X words is supposed to be a proxy for do enough research that you have something to say with depth. A history of the world in 15 minutes better cover enough ground to be worth 15 minutes - as opposed to 1 minute and then filler words. Of course filler is something everyone who writes such a thing and comes up a few words short does - but you are supposed to go find something more to say.
I eventually flipped from moaning about word count minimums to whining about conference page limits but it took a long, long time- well into grad school. The change came when I finally had something to say.
When I was a lazy kid, and there was a requirement to fill a number of lines/pages, I'd just write with bigger letters.
Write a comment explaining that the ostensibly simple task of writing a dozen or so thank you letters for those socks/etc you received for Christmas can, for some people, be an excruciating task that takes weeks to complete, but with the aid of LLMs can easily be done in an hour.
On second thought, you're right. That was easy.
That's old world stuff - I've never sent a thank you card and have lost no sleep bc of it.
I’d prefer to receive no thank you than to receive an AI written one. One says you don’t care, the other says you don’t care but also want to deceive me.
There's the third case: they care. In which you wouldn't be able to tell whether the card is "genuine" or AI written; the two things aren't even meaningfully different in this scenario.
You can tell if the thank-you card is hand-written. Most people don't have a pen plotter connected to their AI text generator to write thank-you notes.
Emailed or texted "thank you" notes don't count. At all.
Yes, but they could also generate the text and transcribe it onto paper by hand.
For many people, myself included, 90%+ work on things like thank-you notes, greetings, invitations, or some types of e-mails, is in coming up with the right words and phrases. LLMs are a great help here, particularly with breaking through the "blank page syndrome".
It's not that different from looking up examples of holiday greetings on-line, or in a book. And the way I feel about it, if an LLM manages to pick just the right words, it's fine to use them - after all, I'm still providing the input, directing the process, and making a choice which parts of output to use and how. Such text is still "mine", even if 90% of it came from GPT-4.
I guess if someone went through the effort to prompt an LLM for a thank-you card note, and then transcribed that by hand to a card and mailed it, that would count. It's somehow more about knowing that they are making some actual effort to send a personalized thank you than it is about who wrote it.
But honestly I don't think "blank page syndrome" is very common for a thank-you card. We're talking about a few sentences expressing appreciation. You don't really have to over-think it. People who don't send thank-yous are mostly just being lazy.
My financial advisor sends out Christmas cards and Birthday cards. They are pre-printed stock cards. I don't even open them. I should tell him not to waste the money. If he even wrote just one sentence that expressed some personal interest, then they would mean something.
These kinds of messages are on the one hand just pro-forma courtesies, but on the other hand they require that that some personal effort is invested, or else they are meaningless.
Sure, it's not just thank you cards though. I once had a job in which my boss assigned me the weekly task of manually emailing an automatically generated report to his boss, and insisted that each email have unique body text amounting to "here's that report you asked for" but stretched into three or four sentences, custom written each week and never repeating. The guy apparently hated to receive automated emails and would supposedly be offended if I copy-pasted the same email every time.
Absolutely senseless work, perfect job for an LLM.
There's something painfully ironic and disturbing that the pseudo-Kolmogorov complexity of clickbait content, as judged "identical in quality" by an average human viewer, is arguably less than the length of the clickbait headline itself, and perhaps even less than the embedding vector of said headline!
It's always been this way, it's just rules of polite/corporate culture don't allow to say what you actually mean - you have to hit the style points and flatter the right people the right way, and otherwise pad the text with noise.
If the spread of AI would make it OK to send prompts instead of generated output, all it would do is to finally allow communicating without all the bullshit.
Related, a paradox of PowerPoint: it may suck as communication tool, but at the same time, most communication would be better off if done in bullet points.
Isn't it just a matter of time until AI gets trained to generate attention-grabbing videos? Also, "the first three seconds" isn't exactly the case anymore. There's a push for algorithm to favour videos that are longer than 1+ minutes. Which, to my understanding, is TikTok's way of fighting for YouTube's userbase.
The videos are longer in total length, but you've never seen the average TikTok/Insta user if you think people are letting videos play for more than a few seconds before scrolling onto the next one. This is why movie trailer videos now have a "trailer for the trailer" in the opening seconds with like "THE TRAILER FOR SONIC 3... STARTS NOW" with all of the most attention-grabbing scenes frontloaded.
Similar on IG, which is why a lot of the photographers on my feed will post five or six images, "Which is your favorite, 1-6? Comment below!" because they get the engagement, synthetic or otherwise, of you clicking through each image, and then commenting.
Youtubers also adopted this trick. The best pattern currently is to voice a random question in the middle of the video or present themselves as "I'm not sure about this, please correct me in the comments" as transparent plan to entice people to comment. I even bought it a first few dozen times, but now when every single creator does it, it is kinda tiring.
I am curious to see this pattern in action. Can you share some @-usernames for these photographers?
> Isn't it just a matter of time until AI gets trained to...
blah blah yes "it's a matter or time" for every one of the myriad shortcomings of the technology to be resolved. If you're a true believer everything is "a matter of time". I'll believe it when I see it.
Depends on the technology. It’s hard to look at the progress from December of 2022 till today, and think we won’t go further. Image generation is getting better every day. Parts of the video generation pipelines are also advancing.
the counterpoint is since then we've killed all the easy ways to scale. the datasets can't get bigger because it's already the whole Internet. model sizes can't grow that much because you start running into RAM constraints. efficiency definitely can be improved, but probably not more than 100x on our current architectures.
I'm also pretty aure you'll see it eventually... Consider the possibility that this is both a bubble waiting to pop, as well as the stuff that will shape the future. Kind of like the Internet around the year 2000.
I don't think this is entirely true. Have you seen these "Internet Story told by TTS Voice-over a Minecraft parkour Video" ( which are what my niblings watching ) ? I noticed a lot of the story is dragged on for over minutes. These are the stories that I read in text in about 5 second.
Short form videos are often hyper focus and to the point, but there are a lot of vertical video contents that are just ( to use the GenZ term ) brain rot like these as well.
Is that true of tiktoks in general? I feel like a lot of the short form videos out there purposefully bait the watcher and drag on for 3/4 of their runtime.
> The younger generation loves short-form, to-the-point stuff.
Meanwhile, Joe Rogan et al. have 3+ hour long podcasts…
Short form video content is absolutely terrible, every time I watch one I have an extremely strong urge to verify whatever I heard in the video, so I just avoid watching them altogether. I’m not sure how anything of value can be 45 seconds long with absolutely no context, outside of comedy.
Counterpoint, GenAI is great at copying styles and typically works best with shorter content.
For example, I could very easily see GenAI being able to produce 1 million TikTok dance challenges.
Which will make them completely worthless by dilution and not stand out. Oops.
Novelty grabs people's attention. A system based on the statistical analysis of past content won't do novelty. This seems like a very basic issue to me.
Novelty itself is easy, the hard part is the kind of novelty that is familiar enough to be engaging while also unusual enough to attract all the people bored by the mainstream.
Worse, as people attempt to automate novelty, they will be (and have been) repeatedly thwarted by the fact that the implicit patterns of the automation system themselves become patterns to be learned and recognised… which is why all modern popular music sounds so similar that this video got made 14 years ago: https://www.youtube.com/watch?v=5pidokakU4I
(This is already a thing with GenAI images made by people who just prompt-and-go, though artists using it as a tool can easily do much better).
But go too soon, be too novel, and you're in something like uncanny valley: When Saint-Saëns' Danse macabre was first performed, it was poorly received by violating then-current expectations, now it's considered a masterpiece.
A system that digs out undiscovered mechanisms to drape novelty on based on where the statistical analysis says it’s already been, that would do it though.
We can sit and imagine horrors the whole week. It has no bearing on the capabilities of machine learning.
If there’s a clear distinction between LLMs looking for patterns in text and LLMs looking for patterns in patterns in text, I’m interested in seeing it better and understanding it.
I ... don't think there's a time limit on TikToks unless you mean that 60 minutes is a time limit. Are you thinking of Vine?
If you want people to watch your content there's definitely a time limit. I don't mean anything imposed by the platform; I just mean if people aren't interested within 3 seconds, they're scrolling to the next video in their feed.
That's a pretty new thing though, in 2020 it was 1 min I believe, and most people skipped after 15-30s
Then it got increased to 3 min and now 10/60 min
Seconds - that's all the time I give something. If it looks draggy I will skip ahead to like halfway thru the video and if the person isn't in the middle of explaining something that they have clearly spent half the time extrapolating, that's it. They get no more time.
I often turn on subtitles and watch at 2x speed.
I prefer the transcript 100% of the time to video.
When I was a kid, I loved having breakfast made for me by other people as I couldnt cook. As a small child, one would say I expected it. It doesn't mean when I become an adult that would continue to be the case.
> If people don't engage with your content in the first 3 seconds, it's worthless.
Is that based on a vigorous experience as a content creator on TikTok as a long form content creator or are you going off what you've heard about TikTok, or it's what your feed is full of? (which says more about you/your feed than it does about TikTok)
https://www.tiktok.com/t/ZTYheCgBq/ 1.1 M views
https://www.tiktok.com/t/ZTYhephfK/ 1.2 M views
or for some more niche stuff which isn't "BIIIG NEWS!! LABOR DAY BREAKFAST GOES HAYWIRE!!!" level of intro:
https://www.tiktok.com/t/ZTYheTVsh/ 54.6 k views
https://www.tiktok.com/t/ZTYheE9m2/
36.3 k views
https://www.tiktok.com/t/ZTYheoK2G/
416 k views
This voice-over definitely isn't going for attention grabbing
https://www.tiktok.com/t/ZTYhe7qTT/ 16 k views
https://www.tiktok.com/t/ZTYhdmJxq/ 64 k views
okay finally found something that's 5 mins long https://www.tiktok.com/t/ZTYhdAfNj/ 270 k views
All of those immediately have something attention-grabbing within the first few seconds: picture of Mars' surface, run map showing funny human shape, text with "Gen Z programmers are crazy" prefacing the anecdote, going immediately into the IT-related rap, guy holding a big and cool-looking stick, "dealing with your 10x coworker" immediately showing the point of the video. The only one that doesn't is the SQL one I guess but that's a very low view count (relatively) on a niche channel.
So thanks for providing a bunch of examples that prove my point, I guess?
none of them are breathless "BIIIG NEWS!! LABOR DAY BREAKFAST GOES HAYWIRE!!!" attention demanding within the first three seconds imo, but, sure, whatever, you're totally right
I've never used tiktok and this post was... enlightening. The ClickUP HR guys are actually pretty funny... but wtf is this from that first channel you posted... lol? https://www.tiktok.com/@luckypassengers/video/74395775119539... - I mean, she's not wrong.
The younger generation has had their brain's development hacked by corporate advertising and can't help themselves but prefer the content they have been all but brainwashed to prefer - this was done deliberately and intentionally and obviously to much detriment to the young of today.
Nobody cares tho bc "they can't pay attention" like it's somehow an entire generation's individual faults they are like that...
Clearly a societal failure there.