Nobody cares about "AI" (Chatbot: I disagree.)

[Pancake]

I wonder how long before people start including instructions like "make a few credible typographical errors, misspelled words, and inappropriate punctuation."

I always include “Do not use — in your response to separate thoughts” since I find they all use this and a sure give away of AI written cover letters and proposals I receive.

 

[Innula Zenovka]

I am so tempted ...

Check out this chat

Here's a chat someone thought you'd want to see.

chatgpt.com

 

[Free]

I am so tempted ...

Check out this chat

Here's a chat someone thought you'd want to see.

chatgpt.com

I'm not sure I get the appeal.

 

[Innula Zenovka]

I'm not sure I get the appeal.

I'm sorry -- I should have expanded on that. As I keep saying, I find AI very helpful for coding, but it's taken me a couple of years to learn how best to use it, and a lot of that has been learning the kind of prompts and settings it requires to make it do what I want it to. I think a lot of people are dissatisfied with AI, or use it to produce results everyone else finds unsatisfactory because, like any other tool, it takes time and practice to understand how to make it work for you.

So if I don't want AI to produce text for me that's obviously written by AI, it seems reasonable that it'll take some effort to make it write the way I want it to, and since AI is, as we are so often told, a stochastic parrot (though I don't think that adequately explains how it can look at my code and tell me why it's not working as expected), trying to get it to write like me rather than a stereotypical AI is probably a good start.

I'm not sure I'll ever use it for anything. The few times I've asked it to write boilerplate emails for me it's produced perfectly serviceable results that have led to useful replies, so I don't really feel the need to improve on that, and I don't want to use it to write posts here, except perhaps when I want to summarise something, but it just seems an interesting, and possibly useful, experiment.

 

[Free]

Still not seeing it. Especially in regards to the "Check out this chat" material, which was about having it write social media posts (for "you", I assume).

Setting up an A.I. agent's parroting skills to use only ones previous writing to train on may mean it could possibly replicate ones style or word choice. But social media content tends to be confused, messy, opinionated and often dealing in current novelty. How could anyone expect it to respond to some topic or event presently in the zeitgeist you've never written on or engaged with previously? As we know it can't think your thoughts, so unless the goal is to have it post only on topics you're already discussed, or stick to simplistic platitudes, it seems like the effort would be pretty limited.

 

[Noodles]

I am so tempted ...

Check out this chat

Here's a chat someone thought you'd want to see.

chatgpt.com

I actually would love to see what came up from a local AI trained on all my writing ever. I have converted essentislly everything I have ever written, including paper joirnals, since the 90s, into Markdown files. I have no idea the word count but its easily tens of thousands of little 500-10,000 word snippets. Probably millions of words.

Honestly, Claude already feels like its picked up a lot of my natural patterns when responding to me. And it already remebers everything about most of the computers I have hat it help me do things on. Often just by computer name, which shows up when I paste some terminal commands in.

 

[Jopsy Pendragon]

I have mixed feelings about this.

Having A.I. write posts as me feels a bit like letting a ventriloquist dummy wear my shirt and improvise opinions. My posts are usually half “I just thought of this,” half “I may be contradicting myself,” with a generous dusting of phone/tablet auto-corruptions I failed to catch. Not sure I need that automated.

But using A.I. as a proofreader / sentence untangler? Sure. Catch the wrong-word substitutions, flag the sentence with six clauses and no safe exits, maybe suggest “this is still you, but less of a hedge maze.”

The trick is not sanding off the fingerprints. Too little help and it’s my usual typo goblin parade. Too much and I start sounding like a corporate apology email.

So, useful? Yes. As a broom, not a ghostwriter. Which up to this sentence was entirely chatgpt responding to a general 'in my voice, what might I say in response to this thread?' It sounds a bit more like my reddit voice than my vvo voice. And having had a chat with chatgpt about this very topic in the past, it's pretty spot on with my opinion.

 

[Kamilah Hauptmann]

I've been using it when some hostile jerk comes at me on Reddit. I dump the screen grab of the context and post in and ask 'what's this jerk's angle'?

4/5 times it turns into a discussion about logical fallacies or something and I lose my desire to fight the jerk.

Sometimes I just post a list of logical fallacies the jerk used and leave jerk twisting in the wind.

 

[Free]

Goats rule. LLMs drool. (Goats drool as well, but not without purpose.)

Frustrated Microsoft Researcher Uses Goats in 'Age of Empires II' to Demo the Absurdity of LLMs

Would you view a GoatGPT response differently to ChatGPT? Even if they worked the exact same way?

gizmodo.com

Perhaps prompted by the galaxy-brained singularity-themed nonsense emerging from his contemporaries at competing companies, a researcher named Adrian de Wynter decided earlier this year to demonstrate that claims for and against the sentience of LLMs require some manner of actually measuring the validity of such claims. In particular, as described in his paper “If LLMs Have Human-Like Attributes, Then So Does Age of Empires II,” de Wynter sat down to demonstrate that at present, we lack any reliable “widely-accepted experimental protocols or schools of thought” for evaluating claims of sentience.

As the title suggests, the paper argues that if LLMs have human-like attributes, then so does the 1999 real-time strategy classic Age of Empires II. But not just any old part of Age of Empires II, mind. No, it’s the goats. De Wynter used the AoE II scenario editor to use the game’s goats as components in basic logic gates. (The details of how he did so are interesting, and use the term “bit-goat”, which we resolve to use as often as possible going forward.)

OK, so what?

This is all fun and games, but what’s the point that de Wynter is making here? There are actually two key points, and they’re both to do with how we go about evaluating an LLM’s anthropomorphic qualities. The first point is that, as demonstrated by the goats, “any sufficiently powerful substrate could implement an entity equivalent to an LLM.”

The term “substrate” is important here, and it basically refers to the “stuff” from which the LLM is built, be it a large codebase stored safely—well, allegedly—at a company like Anthropic or Open AI, or a bunch of virtual goats in AoE II.

The second point, and arguably the more important one, is that “said implementation alters the representation of an LLM, and thus could affect its perceived properties.” Essentially, you could build the same LLM on different substrates, in the same way that you can run the same program on different operating systems.

However, in the case of an LLM—and, specifically, in the case of trying to evaluate that LLMs’ anthropomorphic qualities—the nature of the substrate affects how the LLM is perceived. Crucially, this occurs regardless of the nature of the assumptions made about the LLM’s qualities: “assuming the existence or non-existence of generalised anthropomorphic attributes in order to test a hypothesis proving or disproving their existence is flawed.”

 

[Jopsy Pendragon]

They could dress AI up like a dancing goat juggling bloody knives.... and I would still trust anything coming out of The White House less than it.