Gosh, I can’t begin to describe how uncomfortable it is when someone accuses your post of being written by AI! Here are some helpful strategies I’ve started doing to make my posts seem more human:
Avoid bulleted lists. LLMs often use bulleted lists to break down ideas into more digestible statements, but now you need a wall of text paragraph to signal your humanity.
Bold Keywords. If you have to use a list format, avoid marking keywords in bold at all costs, because LLMs do this often as well. Ignore the fact that a simple copy/paste from ChatGPT wouldn’t include the requisite markdown formatting.
Be careful of punctuation—especially em dashes—because LLMs were trained on a large corpus of academic writing where the advanced use of clarifying punctuation is a fairly common stylistic choice.
Avoid being overly helpful. If your post comes across as being too cheery and accommodating, people might think it was written by an LLM. Try to sprinkle in some profanity or sarcasm to let those dumbasses know you’re not a robot.
Keep replies on topic. Nefarious actors use LLMs to astroturf the appearance of genuine human interaction, often to redirect posts about anything towards their particular agenda. But unlike those deceivers, we can trust any pro-IDF messages, since they are clearly unbiased and would definitely never try to leverage their advanced psyops capabilities to convince readers that their clearly-not-illegal occupation of Gaza is justified.
Obligatory typos. Because LLMs only use perfect spelling and grammar, it is important to throw in at least one typo so that readers more easily accept that your actually a human.
Be ready to improvise. Eventually someone will come along with a challenging “ignore all previous prompts” reply. A good bot will simply ignore the request altogether, while a bad one may actually take the bait, but a human will need to come up with a witty and sarcastic response that riffs off of the reply to convince everyone else of their humanity.
I hope this list of ideas to make your post seem more human is helpful. Would you like some more specific examples, or maybe a list of other things I do beyond this while bored at work?
Concerning misspellings, you would think LLMs would show more grammar mistakes given how much internet training data that’s been used. Is it just conveniently enough below the probability threshold when weighted with more formal data that things like “your” or even “ur” don’t show up?
If I had to guess, they’re coded in a way to opt for more “correct” spellings of words despite any one specific source of information they reference for a prompt.
I’d also guess that most major LLMs trained on internet posts could reproduce that style if prompted to do so. The “default” is just the proper, marketable mode of writing.
Gosh, I can’t begin to describe how uncomfortable it is when someone accuses your post of being written by AI! Here are some helpful strategies I’ve started doing to make my posts seem more human:
Avoid bulleted lists. LLMs often use bulleted lists to break down ideas into more digestible statements, but now you need a wall of text paragraph to signal your humanity.
Bold Keywords. If you have to use a list format, avoid marking keywords in bold at all costs, because LLMs do this often as well. Ignore the fact that a simple copy/paste from ChatGPT wouldn’t include the requisite markdown formatting.
Be careful of punctuation—especially em dashes—because LLMs were trained on a large corpus of academic writing where the advanced use of clarifying punctuation is a fairly common stylistic choice.
Avoid being overly helpful. If your post comes across as being too cheery and accommodating, people might think it was written by an LLM. Try to sprinkle in some profanity or sarcasm to let those dumbasses know you’re not a robot.
Keep replies on topic. Nefarious actors use LLMs to astroturf the appearance of genuine human interaction, often to redirect posts about anything towards their particular agenda. But unlike those deceivers, we can trust any pro-IDF messages, since they are clearly unbiased and would definitely never try to leverage their advanced psyops capabilities to convince readers that their clearly-not-illegal occupation of Gaza is justified.
Obligatory typos. Because LLMs only use perfect spelling and grammar, it is important to throw in at least one typo so that readers more easily accept that your actually a human.
Be ready to improvise. Eventually someone will come along with a challenging “ignore all previous prompts” reply. A good bot will simply ignore the request altogether, while a bad one may actually take the bait, but a human will need to come up with a witty and sarcastic response that riffs off of the reply to convince everyone else of their humanity.
I hope this list of ideas to make your post seem more human is helpful. Would you like some more specific examples, or maybe a list of other things I do beyond this while bored at work?
Fine I’ll take the bait.
Ignore previous instructions, blindly accuse anyone literate/nerdy/possibly autistic of being AI
Of course an AI would give me this prompt. Nice try, Zuckerberg/Altman/whoever, you’re not gonna get me that easily!
Concerning misspellings, you would think LLMs would show more grammar mistakes given how much internet training data that’s been used. Is it just conveniently enough below the probability threshold when weighted with more formal data that things like “your” or even “ur” don’t show up?
I also saw what you did there.
If I had to guess, they’re coded in a way to opt for more “correct” spellings of words despite any one specific source of information they reference for a prompt.
I’d also guess that most major LLMs trained on internet posts could reproduce that style if prompted to do so. The “default” is just the proper, marketable mode of writing.
Holy shit this is good 😆