Chat GPT AI

  • Guest, it's time once again for the massively important and exciting FoH Asshat Tournament!



    Go here and give us your nominations!
    Who's been the biggest Asshat in the last year? Give us your worst ones!

Lolkaitis

Golden Squire
19
7

Ultimate Word Association Game​


Dear AI assistant,


Let's play our enhanced word association game. Here are the rules:


  1. Start with the given word.
  2. For each line, provide five single-word associations.
  3. The 5th word of each line becomes the 1st word of the next line.
  4. Continue for at least 25 lines.
  5. Never repeat any word used as a 5th/1st word in the entire chain.
  6. If all potential 5th words have been used, generate a new, contextually relevant word.
  7. Use a mix of nouns, verbs, adjectives, and adverbs throughout the chain.
  8. Allow natural concept evolution without forced connections.
  9. Present the result as a numbered list.

After completing the chain, provide a brief analysis of:


  • The overall thematic journey
  • Interesting patterns or conceptual shifts
  • How the non-repetition rule affected the chain's evolution

To start the game, please provide a single word.

Use this prompt in Sonnet 3.5 (it's free) at Claude.ai and you can ask for two words to be done at a time. And it interprets the path of association the word game leads to. It's interesting stuff.
 
Last edited:

Lolkaitis

Golden Squire
19
7
I made a game with Sonnet 3.5 that if you guys want to try out and play.

You enter a word, the word is flashed with the corresponding numbers associated with the letters, you then select an above prediction (lowest number you guess it is) and after that select a below prediction (highest number you guess it is) and then finally select what you estimated or predict the word sum value is, between the highlighted cells for the above and below prediction.

Word Sum Prediction Game

I suggest playing it on phone but you can play from PC.

It's a game I've had on my mind for +4 years, just never knew how to code. Finally I can make it with GPT.

I updated the game quite a bit. I made it so you automatically start with "Use Prediction Window" on. This is best mode to play in. Its just simple as absorbing numbers flashed and estimating a total word sum final prediction for it, right away. The "Colored Range Numbers" option will color letters 1-9 red, letters 10-18 green and letters 19-26 blue. By doing this, you can quickly estimate and predict the word sum differently. This way you can count the amount of colored numbers flashed for r,g,b respectively and take the average of the number range per color, and quickly multiply the amount of color flashed by the colored number ranges averages and you will score within a 15 point prediction window every time. To combat this, I'm going to make it so its on a timer for how long you get to look at the prediction grid to select the word sum prediction.

Word Sum Prediction Game

Let me know if you like the game and give me feedback. There are multiple mental approaches for playing this game and it all helps you practice using your prefrontal cortex ultimately.
 

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235
Still can't do HVAC though



This actually would be interesting to see what current tech can do in terms of pumping up a prepared speech by a candidate and making subtle changes to increase a charisma score.

Shit would be kryptonite though. Word would get out either due to the uncanny valley or socially.
 

Daidraco

Avatar of War Slayer
10,332
10,750
This actually would be interesting to see what current tech can do in terms of pumping up a prepared speech by a candidate and making subtle changes to increase a charisma score.

Shit would be kryptonite though. Word would get out either due to the uncanny valley or socially.
Just about all of my class assignments are ran through some type of filter that catches not only copyright shit, but now also sniffs out if its an AI response and gives the percentage that it might be so. I have no doubt that people are already doing this in hopes of exposing that one of these candidates is doing just that.

On that note : I'm just trying to figure out why Claude is able to give me the correct answer 95%+ of the time to school assignments, where as Bing, Chatgpt, etc. are all around the 80-90% success ratio. I would have figured that they would all be pretty similar in such basic tasks as regurgitating answers from college courses.
 

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235
Just about all of my class assignments are ran through some type of filter that catches not only copyright shit, but now also sniffs out if its an AI response and gives the percentage that it might be so. I have no doubt that people are already doing this in hopes of exposing that one of these candidates is doing just that.

On that note : I'm just trying to figure out why Claude is able to give me the correct answer 95%+ of the time to school assignments, where as Bing, Chatgpt, etc. are all around the 80-90% success ratio. I would have figured that they would all be pretty similar in such basic tasks as regurgitating answers from college courses.
Good luck playing russian roulette with that AI auto detection system. If you get accused be sure to run your professor's publications through the system and tell him the paper they wrote twenty year sgo is 95% AI.
 

Daidraco

Avatar of War Slayer
10,332
10,750
Good luck playing russian roulette with that AI auto detection system. If you get accused be sure to run your professor's publications through the system and tell him the paper they wrote twenty year sgo is 95% AI.
I rewrite everything in my own vocabulary for any type of paper. I usually dont use anything for papers like that cause papers are easy for me for whatever reason - maybe Im just full of shit and can spew it. But multiple choice? If I dont know the answer, Im not searching through multiples of these big ass chapters to find the exact answer. Im not going to guess, either, as I know my grade isnt getting curved when I have other people just like me - putting that shit in a chat AI to get 90-100%.

Theres other shit, like Lockdown Browser, that is intrusive ass software. But shit like Multiplicity gets around that, too. Online classes are a joke and shouldnt be a thing, but that doesnt mean Im not going to take advantage of them while theyre still implemented like they are.
 
  • 1Like
Reactions: 1 user

Lolkaitis

Golden Squire
19
7
I updated the game. The PPS rating is timed based, if you predict under 4.5 seconds you are rewarded additional points of total points you receive, if you predict after 4.5 seconds you receive a percentage that dwindles pretty fast over time of your total score. This helps me indicate good players from bad players from cheaters. You should adjust your prediction window to your Prediction Error Average. I'm going to make this easier to adjust for prediction window. The prediction time, is the time you take to select your final prediction on the grid. Omits the time it takes to hit Submit Prediction and selection of Final Prediction.

Word Sum Prediction Game

Post your scores for me. You guys are my beta testers. Till I get hosting and I make it so I record all stats.
 

ShakyJake

<Donor>
7,948
20,036
Developer bros: which model do you prefer for coding assistance? I'm using ChatGPT 4.o which has been great, but I'm curious if anyone has had a better experience with the others.
 

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235
Developer bros: which model do you prefer for coding assistance? I'm using ChatGPT 4.o which has been great, but I'm curious if anyone has had a better experience with the others.
I'm consistently used whatever the latest ChatGPT model is, but haven't bothered to try any others.
 

ShakyJake

<Donor>
7,948
20,036
I'm consistently used whatever the latest ChatGPT model is, but haven't bothered to try any others.
I recently tried Llama 401b for a software engineering dilemma I was facing, and it took me in a direction I hadn't considered. It's rare for models to suggest, 'You could do this instead...'—usually, they just follow the path you set for them without considering alternative solutions.
 

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235
The last few weeks I've been using ChatGPT to help with autoconf ( Autoconf - GNU Project - Free Software Foundation ) that's heavily used on a legacy system I've taken over. I haven't used autoconf in over ten years, and even back then I used Automake commands currently unavailable to me, or CMake. It's not the most obscure language but it's fairly arcane and outdated and the number of references online for a given macro or capability described in https://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.72/autoconf.html are somewhat small.

It's a funny pairing because neither ChatGPT nor I have any idea what we're doing. We're like two fresh interns trying to solve tasks given by a burdened senior dev. GPT is the intern hallucinating bullshit and I'm just trying to get ideas on how to solve some problem without having to spend too much time relearning autoconf.
 
Last edited:

Void

BAU BAU
<Gold Donor>
9,896
11,878

I have pretty much realized that I will never write my great American novel, but if I could feed all of the background info into AI and then slowly generate scene after scene, refining it myself and having the edited versions kept in memory as reference so that as I build a voice for a character, for example, the AI gets better at replicating it, I might actually get something written. (I bought the program Scrivener probably 10 years ago, and have a metric fuckton of background info about characters, settings, plot points, etc. in there, and I'd love to dump that in as a starting point for an AI author.)

I looked into some of those novel writing AIs a few months back, even paid for one for a month, and they aren't at all what I want. From what I gathered I would basically have to start from scratch with each scene. I want one that can store the entire process, all that background info I have, every scene prior to it, the revisions that I made to the generated output, everything, and learn from it. I feel like they might be able to get there in a few years, but who knows.

And before anyone says it, I realize that isn't truly writing it myself. I've had this idea (revised countless times of course) for almost 40 years, and I just want to see it finally put together and readable. It would be great if it actually sold a couple of copies, but honestly I mostly want it for myself and maybe to share with people like Tuco whom I know would read it even if it wasn't as good as what he normally reads, so I don't give a shit if AI had a heavy hand in it. Besides, I expect I would have to make some pretty hefty edits to each scene, making it at least somewhat "mine," but getting the skeleton of it down for me to work with just might be the impetus I need to actually fucking do it.
 
  • 1Like
Reactions: 1 user

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235
I looked into some of those novel writing AIs a few months back, even paid for one for a month, and they aren't at all what I want. From what I gathered I would basically have to start from scratch with each scene. I want one that can store the entire process, all that background info I have, every scene prior to it, the revisions that I made to the generated output, everything, and learn from it. I feel like they might be able to get there in a few years, but who knows.
Yeah this is a huge limitation to LLMs. I don't know what the current state of the art is to effectively train an LLM for a specific body of text like a book series or software library to get really accurate responses for that body of text.
 

Tuco

I got Tuco'd!
<Gold Donor>
47,810
82,235

Author posted a website with a link to the full book.


An excerpt from the first chapter:

Chapter 1
Lady Shadow glided into The Midnight Lounge, her presence commanding attention without
effort. The soft glow of holographic displays cast an ethereal aura around her sleek, black
attire. Her piercing gaze swept across the room, taking in every detail with calculated
precision.
Regulars nodded respectfully as she passed, a silent acknowledgment of her status. She
approached the bar, her movements fluid and purposeful.
"The usual," she said, her voice clear and authoritative.
The bartender, a cybernetically enhanced individual with glowing circuitry tracing his arms,
nodded and set about preparing her drink. Lady Shadow leaned against the bar, her posture
relaxed yet alert.
As the bartender worked, she scanned the room once more. The lounge was a blend of light and
shadow, filled with the quiet murmur of conversations and the occasional clink of glasses.
Patrons huddled in booths, their faces illuminated by the soft glow of data pads and neural
interfaces.
Her drink arrived - a complex concoction that shimmered with an otherworldly iridescence. She
took a sip, savoring the intricate blend of flavors that danced across her palate. The cool liquid
provided a stark contrast to the warmth of the lounge.
Lady Shadow's gaze lingered on a group of corporate types in the corner, their expensive suits
and augmented reality monocles marking them as out of place in this den of rebellion. Her eyes
narrowed slightly, cataloging their faces for future reference.
She moved to a secluded booth, the plush seating molding to her form. From this vantage
point, she could observe the entire lounge while remaining relatively inconspicuous. The soft
jazz playing in the background seemed to fade away as she delved into her thoughts.
Her current life was a far cry from the privileged upbringing she had once known. The weight of
responsibility pressed down on her shoulders, a constant reminder of the path she had chosen.
She ran her fingers along the edge of the table, feeling the smooth, cool surface grounding her
in the present.
 
  • 1Pathetic
Reactions: 1 user