AI-generated writing

NC: I’ve experimented with MidJourney to make images but how is the world of AI-generated text going? I’m trying out the LEX, a cross between a Google docs wordprocessor and an AI text generator. This paragraph is actually me typing but the next one…well that will be up to the machine

[AI] LEX is a great tool for writers who want to get their ideas down quickly and don’t have time to worry about grammar or punctuation. It’s also great for writers who want to experiment with different voices or styles.

[NC] OK that was going to be a bit easy for it. Just a chunk of promotional text. What I need to see is if it can come up with something a bit more off the wall. Time to introduce Timothy into the mix. Let’s see what happens if I request some info about the world-famous Timothy the Talking Cat.

[AI] Timothy the Talking Cat is a world-famous feline who has his own line of products and a popular YouTube channel.

[NC] I wouldn’t call his YouTube channel “popular” but OK, not bad. Less background. What can it tell me about Susan the Triceratops?

[AI] Susan the Triceratops is a prehistoric creature that lived during the Cretaceous period.

[NC] And Mr Atomic the Robot cleaner?

[AI] Mr Atomic the Robot cleaner is a robotic vacuum cleaner that is designed to clean your home.

[NC] Can this AI handle more complex topics? What I’d like is a summary of the Sad Puppy Debarkle.

[AI] Unfortunately, the Sad Puppy Debarkle is a complex topic that is difficult to summarize.

[NC] No worries. How about a lengthy explanation of the Sad Puppy Debarkle?

[AI] The Sad Puppy Debarkle is a complex topic that is difficult to explain.

[NC] Plato had some fascinating ideas about philosophy.

[AI] Plato was a Greek philosopher who is considered to be one of the most important thinkers in Western philosophy. Plato’s philosophy is characterized by its emphasis on ethics, justice, and what he called the “Forms” (idealized versions of objects or concepts).

[NC] So the sentences it is generating are reasonable. I assume it’s pulling this from web searches. However, I’m struggling to see how this will be useful.

[AI] The LEX can be useful for writers who want to get their ideas down quickly and don’t have time to worry about grammar or punctuation. It’s also great for writers who want to experiment with different voices or styles.

[NC] That’s the official line but I’m struggling to see how.

[AI] There are a few potential benefits to using the LEX. For one, it can help you get your ideas down quickly without worrying about grammar or punctuation. This can be especially helpful if you’re trying to capture a stream of consciousness or brainstorming session. Additionally, the LEX can be helpful in experimenting with different voices or styles. This can be a fun way to play around with your writing and explore new possibilities.

[NC] I really don’t see this helping with a stream-of-consciousness style of writing or with not worrying about grammar etc. I’m stopping every so often to prompt the AI into producing something, which isn’t a great way of having a flow.

[AI] The LEX can be a great tool for writers who want to get their ideas down quickly and don’t have time to worry about grammar or punctuation. It’s also great for writers who want to experiment with different voices or styles.

[NC] It’s stuck on that topic now. Let’s redirect. Is a picture of a triangle actually a triangle or is it merely a semblance of a triangle?

[AI] The answer to this question is complex and dependent on your perspective.

[NC] Question marks might cue it to treat text as a question. When I once suggested that a picture of a triangle is not the actual ideal of a triangle I helped provoke a really weird right-wing debate tactic called “the witch test”.

[AI] The “witch test” is a controversial debate tactic that has been used by right-wing thinkers to try and disprove the existence of certain concepts or ideas. The basic premise is to take a picture of a purported witch, or some other creature that is not actually real, and then ask if the person looking at the picture believes that it is a real witch. If they do, then they are accused of being gullible or foolish. The tactic

[NC] That was cool!

[AI] The LEX can be a great tool for writers who want to get their ideas down quickly and don’t have time to worry about grammar or punctuation. It’s also great for writers who want to experiment with different voices or styles.

[NC] Darn. Back to the witch test. I liked how it had the start of a sentence for me to continue. OK. The tactic could also involve a picture of a triangle but this initially confused the right-winger to imagine they saw demons.

[AI] The “witch test” is a controversial debate tactic that has been used by right-wing thinkers to try and disprove the existence of certain concepts or ideas. The basic premise is to take a picture of a purported witch, or some other creature that is not actually real, and then ask if the person looking at the picture believes that it is a real witch. If they do, then they are accused of being gullible or foolish. The tactic

[NC] The tactic can also include banana bread.

[AI] The “witch test” is a controversial debate tactic that has been used by right-wing thinkers to try and disprove the existence of certain concepts or ideas. The basic premise is to take a picture of a purported witch, or some other creature that is not actually real, and then ask if the person looking at the picture believes that it is a real witch. If they do, then they are accused of being gullible or foolish. The tactic can also include banana

[NC] OK, that’s enough for now 🙂 Goodnight

[AI] Thank you for using the LEX. Goodnight!

,

14 responses to “AI-generated writing”

  1. That seems like a disimprovement on Eliza. And I had heard that AI was writing fake scientific papers. (Though perhaps only predatory journals would publish them.)

    Like

  2. That was amusing but fairly useless. The witch test sentences were good, but Puppies aren’t complex at the roots of it all. Mostly it seems to be advertising itself!

    Liked by 1 person

  3. A lot of AI writing is like that: mostly slightly boring and what you expect, but then every so often you get a result that makes you think ‘what dark part of your training set did that come from?’ I once fed GPT3 the prompt ‘[my son’s name] should do his homework’ and one of the sample paragraphs it generated ended up complaining about how hard it was to talk to girls.

    Liked by 2 people

  4. When I was 18, just before I left for college, I met the new physics teacher at my high school, who’d gotten his degree from MIT. He showed me a script from the DOCTOR program, much of which I remember to this day. (Or think I do.) It illustrates how the funniest exchanges require a clever (if not malicious) user.

    Help me, doctor! I’m afraid of the dark.

    You’re afraid of the dark.

    Yes, and if you don’t help me, it may cost me my wife.

    Tell me more about your family.

    It’s just me and my wife still, because I’m afraid of the dark.

    You and your wife still?

    No! My wife and I don’t still, damn it!

    Are you sure you’re not in denial?

    [Change of human operator here]

    You’re right. My wife is unhappy because when I get really afraid of the dark I booze it up.

    You booze it up?

    Yes. Many’s the time she’s found me laying on the floor passed out drunk because I was afraid of the dark.

    Maybe many really is the time you was laying on the floor passed out drunk because you was afraid of the dark.

    Yes, doctor. This is true.

    For years after that, whenever anyone would try to talk seriously about any sort of AI conversation engine (not counting practical dialogue systems), I’d reply with something like, “Maybe many really is the time.”

    (But then people started to think I was strange . . .) 🙂

    Liked by 2 people

    • I’ve seen some quite effective website-help chatbots at work these days. Obviously, a lot of that depends on the quality of background material people are adding to it though.

      Like

  5. We tested an AI text system at work recently. It was trained on the text we wrote in the customer service system, and was supposed to help us write responses by suggesting the next sentence.

    Around 50 % of the time, the suggestion it made was the signature of a coworker.
    “Kind regards
    (name of coworker)
    Customer service representative”

    It was almost never useful. It was frequently in the neighborhood of correct, but those where cases where I didn’t have to think very much to write a reply. And it got the grammar wrong often enough that I could never trust it – I had to read through the suggestion and make corrections to some words. And that ended up being more work than writing the same reply unassisted.

    Liked by 2 people

Blog at WordPress.com.