LLM writing often puffs up the importance of the subject matter with reminders that it represents or contributes to a broader topic.
LLMs have serious problems keeping a neutral tone
LLMs often introduce their own interpretation, analysis, and opinions in their writing
Conjunctions - Words to watch: on the other hand, moreover, in addition, furthermore …
Parallel constructions involving “not”, “but”, or “however”
LLMs overuse the ‘rule of three’—“the good, the bad, and the ugly”. This can take different forms from “adjective, adjective, adjective” to “short phrase, short phrase, and short phrase”
… there were more like the overuse of the em-dash,
I’m always amazed, surprised and confused when some people think it’s AI but it’s just a well written piece. I’ll assume it’s from warping prose into text / IM messages in addition to less structure in things like email.
I think that writing and writing well are super important, even though video forms like Youtube and Tik-tok are popular. I’m always looking at the footnotes in Youtube technical presentations to get additional details.
However, things do change across time, but as AI gets better and human writing skills decline, there will be a definite blurring.
In closing, I wonder how much of my posts, post from any of the long time forum participants, or coaches / EP in general would pop up as “90%” AI.
I have nothing against writing like AI, but the gripe I have is when people dont put effort into their question and expect people to put effort into their answer.
If I ask ChatGPT “make me an incredibly detailed set of questions about robot design” and it comes up with 10 different topics that I copy and paste into the forums, then I would have put in 1 minute of effort for about 50 minutes of community effort of response.
I would not mind AI if it required the user to put in effort 99% of the time, but usually what it actually is is a way of turning one sentance into many for the sake of seeming more professional.
I have had many conversations with friends and people online about how their writing could be misinterpreted as AI because of advanced grammar techniques that it uses. However, when you actually read what AI says the concepts it applies to the grammar is totally different. I have an example directly infront of me, as grammarly is trying to use its AI grammar tool to completely change every adjective and noun I’m using because it doesn’t understand context. Writing like AI is tremendously difficult, as it requires a writing style that uses
The Grammar Techniques you listed (repeated in almost every sentence)
Bold words and headings
Random Emojis
Avoidance of repeating words related to key concepts
Clear sentences that avoid nuance
Misinterpretation of context
I agree that people shouldn’t be throwing around ‘AI text’ when something is written well, but actual AI writing and scripts are easy to spot, and often cause confusion and honestly show some laziness.
Again, one sentence from a human explaining a problem is 10x more useful than an AI paragraph or writing like AI, morally and for understanding.
Indicators of AI writing such as overusing boldface/em dashes/decorational emojis aren’t my issue with the writing - it’s that they often indicate deeper conceptual issues with the text resulting from the use of AI and a lack of investment in time and thinking from the writer.
These days, even if you can’t prove to others that your writing is not AI-generated, the best thing to keep in mind is that someone claiming your work to be AI-made boils down to a compliment. Sure, it might mean they lack confidence in your writing capabilities, but it does ultimately mean that they are impressed (whether they admit it or not).
In a way, this is similar to all of the recent discussions regarding rule changes to crack down on mentor-built, copied, or even purchased robots. No rule or regulation will ever be 100% foolproof, but the one thing that is present for every instance of someone cheating or using unintentional loopholes to work around rules is guilt. It’s not always large, but there is always at least a little bit of regret.
At the end of the day, the true winners of the competition are those who grew as people, engineers, had a fun time, etc. The best way to do these things is to be honest with yourself and hold yourself to good standards.
So… all of us must remember that no matter what is going on, what others think of us isn’t our absolute first priority. Doing what’s right (and knowing that our work is good) is.
I am tired of AI generated troubleshooting web pages and responses from Google. The instructions can not only be misleading but can actually damage the computer, especially for Linux systems.
Immoral use of AI sucks. And AI detection software sucks more. Making a system that hallucinates grade if there is a system that hallucinates can cause hallucinations. AI detection software will always be indefinitely behind. It’s a catch 22. The easiest solution is to not try to beat oneself with desciphering between AI and human, but provide more logically complex questions, which would usually trip up the AI explained in my point “5.” below. The AI generated response would immediately receive a low score through traditional human grading.
You can talk to a new member from a help desk who has more experience and specialization. However, for an AI assistant, a LLM is a singular model. You can only get help based upon the pre-existing trained billions of dials for that one model version. So there is very very high intertia between the quality of most recent data when comparing someone with genuine experience versus an LLM.
AI Generated videos, TTS narration, etc. Is fatiguing on YouTube. And the worst part is that the narration is awful and can be misleading presenting incorrect statements as factual claims.
AI is highly shallow. The way that it has improved is by adding tensors on top of tensors and for tensors. Therefore, more sophisticated responses (abstract and/or logical) can start to be achieved but not appropriately or with thought. It is still with regurgitative capacity. I personally think if OpenAI paired with WolframAlpha, ChatGPT would have a lot more holes covered with its mathematical computational solving capacities.
As for the original post…
In the past I used Grammarly, which is a grammar helping tool, well before the AI craze of 2020. However, if I use it now, many submission systems have a high risk of flagging my work as AI generated because it’s ‘too grammatically sterile.’ It is exceptionally annoying because I wrote it in my own terms. So, I just let Grammarly do it’s thing and accept the annoying risk of having certain work flagged as AI generated even though it is clearly not.
Mavis Beacon Teaches Typing and Grammatik were two software suggestions I made to people in the 1980’s I was lucky, I had access to the Bell Labs Writer’s Workbench.
Grammarly is a great add on, I don’t know why more people don’t use it.
And I commiserate on the AI slop that now is the help desks and a huge number of AI youtube videos.
“Don’t you know how important I think I am?” A good chunk of what VRC/IQ/etc. is about is communications. I found that the best communicators get their ideas out there more often than not.
I’ve been avoiding this topic like a landmine but I feel like I gotta get something out of my mouth.
I get the whole ‘strive to write like AI’ thing- but also I hate it now. I’ve been called out for using AI so many times and it doesn’t really seem to get better. And it’s not even in robotics- It’s also in English- And science!
I honestly wanna steer away from writing like AI! I try to change my tones when writing to sound more ‘human’ or something but then when I get to trying to do well on my work I always go back to those old habits!
I used to have a part of my notebook explaining the design process (Replaced with this: The design process is a process followed by engineers and other STEM based jobs when creating an solution to a problem. This year we, as a team, will work on trying to find a solution to the VEX V5 game “Push back”
Now, we’ve said that we’re going to find a solution to Push Back, but… What is considered a solution?)
I still remember how my team members looked up at me (due to the old wording I once had; now reworded due to the updates in my notebook, and said “I thought we’re not supposed to use AI? Does this mean that I can use ai for ____”
Pity, pity.
I don’t think anybody should strive to write like AI, maybe just try to write decently.
And maybe people need to tone down with AI use… And AI accusations.
And for the sake of this conversation to keep on going (cause it’s likely gonna be the biggest thing since that G4 violations topic)
Was like SOOO long ago. (Back when I was newer to the vex forums and weird and all that, hahahah)
Anyways, going to test that statement I made so long ago.
All of this “AI” garbage has been bugging me for a long time. Since the end of my 8th grade year, or thereabouts, I’ve had multiple essays I wrote — entirely by myself, mind you — flagged as yet another heap of steaming horse manure ChatGPT spat out. As this tech has developed, I find myself increasingly frightened by its proliferation in today’s society. I take a certain stance (and despite how this comic presents it, it’s a valid concern) on LLMs: USE THEM FOR THE BUSY WORK AND NOTHING ELSE.
It’s concerning how few people utilize these systems in ethical and reasonable ways. I consider myself to be an artist and a writer — among other things — and those parts of me feel the most threatened by them. The fairly common use of LLMs as book generators and art machines (among otherthings) is a depressing reminder of how eager people are to throw whatever they can think of at new tech. The whole industry shamelessly rips from thousands — millions, even — of books, images, social media posts, and for what? To write an essay for a high schooler? To crank out generic, clichéd novels to sell on Amazon?
I don’t support their use for any reasons other than coding (with revisions) or intentionally breaking them.
That’s my two cents. I’ll get off my soapbox now.
A good chunk of what VRC/IQ/etc. is about is communications. I found that the best communicators get their ideas out there more often than not.
