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!

Mist

REEEEeyore
<Rickshaw Potatoes>
31,797
24,465
While we're talking about the inevitable cognitive decline that results from outsourcing cognitive functions to the machines

Gen Z started off dumb already but holy shit now they are fucked
Did you know that 99.89% of people google shit at work?

I'm skeptical of AI, and I worry about kids using it in school instead of learning, but it just isn't good enough to do real work for you. It's like having a moral panic that people are posing questions on stackexchange at work.
 
  • 1Worf
Reactions: 1 user

Bandwagon

Kolohe
<Silver Donator>
24,984
68,541
I wrote my first screenplay. Only took 5 minutes.

Title: Cajun Taters
A Bayou Comedy of Errors


Logline:
Channing Tatum is sent deep into the Louisiana bayou to prepare for his role as Gambit in the MCU. His agent, played by Keegan-Michael Key, sets him up to live with a no-nonsense Cajun named Taters Chanel (also played by Tatum). Over three months, the duo embarks on a series of wild, politically incorrect, and often dangerous misadventures as Tatum struggles to adapt to bayou life, all while trying to master Gambit's accent and flair.




ACT 1: Fish Out of Water
Hollywood pretty boy Channing Tatum is riding high after being cast as Gambit, but his agent, Reggie (Key), insists that he needs to “authentically” embrace Cajun culture to nail the role. Reggie arranges for Tatum to stay with Taters Chanel, a grizzled, half-mad, possibly inbred Cajun hermit. Upon arrival, Tatum is immediately thrown into situations he never anticipated—sleeping in a shack on stilts, dodging gators, and eating food that’s either fried, fermented, or still moving. Taters, unimpressed with Tatum’s celebrity status, subjects him to a brutal crash course in Cajun life, including an initiation involving moonshine, a stolen airboat, and a questionable wrestling match with a nutria.




ACT 2: The Learning Curve
Tatum slowly adjusts to bayou life, but his attempts at learning the accent and mannerisms are disastrous. Taters insists on practical lessons, including backwoods gambling, poaching, and running various hustles in nearby towns. Each scheme leads to more chaos—whether it’s accidentally robbing a gas station (they thought it was a themed escape room), wrestling a literal bear at a backwater casino, or getting mixed up with a swamp crime lord known only as “Big Pappy.”


Reggie occasionally checks in, horrified at Tatum’s worsening hygiene and growing affinity for wearing overalls without a shirt. Despite the insanity, Tatum starts to respect Taters' way of life. But things take a turn when they stumble upon a local politician’s illegal gator-fighting ring, leading to a high-stakes showdown involving dynamite, a boat chase, and a very confused National Guard unit.




ACT 3: The Gambit Pays Off
As the three months come to an end, Tatum has transformed. He’s nailed the accent (sort of), picked up questionable survival skills, and developed a bizarre but genuine friendship with Taters. However, just as he’s about to leave, Big Pappy resurfaces, blaming them for his business falling apart. A final, absurdly over-the-top bayou battle ensues—airboats, homemade explosives, and a dramatic slow-motion alligator attack.


Tatum ultimately survives, barely, and returns to Hollywood with a newfound confidence and a questionable felony record. Reggie is thrilled to have him back, though horrified at his new pet possum and the distinct scent of swamp water that follows him everywhere.


As the credits roll, we see a clip of Tatum’s final Gambit screen test. He nails the accent, winks at the camera... and then accidentally blows up the entire set with an ill-advised card trick.


FADE TO BLACK.
 
  • 1Like
  • 1Worf
Reactions: 1 users

Mist

REEEEeyore
<Rickshaw Potatoes>
31,797
24,465
Grok 3's summary of twitter discussions of Grok 3:

1739990173057.png
 

LiquidDeath

Magnus Deadlift the Fucktiger
5,121
12,323
I just tried to use Grok 3 to do some simple family tree building using the memorial pages on Find A Grave and it is shockingly bad. I mean, I can't even tell where it is pulling data from bad. All the data and links it needs are right there on each page, birth/death day and parents grave site memorial links under a heading called "Parents, but it just refuses to use these very simple and obvious items. If this is the peak of AI right now, I am thoroughly unimpressed.
 

Chris

Potato del Grande
19,959
-10,417
I just tried to use Grok 3 to do some simple family tree building using the memorial pages on Find A Grave and it is shockingly bad. I mean, I can't even tell where it is pulling data from bad. All the data and links it needs are right there on each page, birth/death day and parents grave site memorial links under a heading called "Parents, but it just refuses to use these very simple and obvious items. If this is the peak of AI right now, I am thoroughly unimpressed.
You need to do that yourself, there's usually too many mistakes, alternate names and common names to be lazy about it.

Ancestry.com uses a mix of AI, genetics and crowd sourcing. It's excellent for doing this at least in the UK which has 200 years of census data.

Even then I'm only trusting genetic matches to build out unknown braches of the family tree.
 

LiquidDeath

Magnus Deadlift the Fucktiger
5,121
12,323
You need to do that yourself, there's usually too many mistakes, alternate names and common names to be lazy about it.

Ancestry.com uses a mix of AI, genetics and crowd sourcing. It's excellent for doing this at least in the UK which has 200 years of census data.

Even then I'm only trusting genetic matches to build out unknown braches of the family tree.

Why would I do it myself when it is literally just following embedded links in a page and creating a simple diagram? Time consuming for a human, incredibly easy for a computer. This should be something that AI excels at, but it is woefully bad. It can't even correctly identify birth/death dates from the base memorial page, both of which are very clearly labeled.
 

Daidraco

Avatar of War Slayer
10,678
11,335
Why would I do it myself when it is literally just following embedded links in a page and creating a simple diagram? Time consuming for a human, incredibly easy for a computer. This should be something that AI excels at, but it is woefully bad. It can't even correctly identify birth/death dates from the base memorial page, both of which are very clearly labeled.
Its not just Grok. Its LLM's in general. They are woefully bad in certain fields that you would "think" would be incredibly simple. But on that same note, where they have a lot of focus - theyre incredibly useful. That airplane game linked here was made entirely by LLM's using a sort of "plugin." The game itself isnt something special, but the fact that it can make a working program with little instruction from the user is pretty impressive.

The thing that I think is hurting the world right now is that each one of these LLM's that comes out - we're ultimately retracing footsteps over and over and over. Training on the same or similar data to accomplish the limited set of things that it can be easily trained upon. We would obviously never agree to do it, but a central LLM that isnt under control of just one particular party would leap frog us several years into the future. Ive read fanfiction or plausible theories? About a LLM working from and being stored on a block chain type of system to where a single company cant program it to be politically aligned in one way or another - but until thats a reality, we're stuck with what we got.
 

Captain Suave

Caesar si viveret, ad remum dareris.
5,605
9,632
Why would I do it myself when it is literally just following embedded links in a page and creating a simple diagram? Time consuming for a human, incredibly easy for a computer. This should be something that AI excels at, but it is woefully bad. It can't even correctly identify birth/death dates from the base memorial page, both of which are very clearly labeled.

This isn't a good use case for how LLMs work right now. Data that were not part of their training set and aren't in the prompt ("go to this link", etc) are dealt with through add-on modules and not well-integrated into a solution.
 

Chris

Potato del Grande
19,959
-10,417
Why would I do it myself when it is literally just following embedded links in a page and creating a simple diagram? Time consuming for a human, incredibly easy for a computer. This should be something that AI excels at, but it is woefully bad. It can't even correctly identify birth/death dates from the base memorial page, both of which are very clearly labeled.
Anything mathematical (eg matching names, locations and dates) needs an algorithm or neural network, which is what ancestry.com does and it's extremely effective. You still need to verify.

Large language models are for creative writing and search engine summaries.

You don't want mistakes in your family tree because someone had the same name and birth date as an ancestor, it happens a lot.

Beyond that, you don't want a garbage family tree because great grandma cheated on great grandad. You need a DNA match, ancestry.com does this.
 

LiquidDeath

Magnus Deadlift the Fucktiger
5,121
12,323
Anything mathematical (eg matching names, locations and dates) needs an algorithm or neural network, which is what ancestry.com does and it's extremely effective. You still need to verify.

Large language models are for creative writing and search engine summaries.

You don't want mistakes in your family tree because someone had the same name and birth date as an ancestor, it happens a lot.

Beyond that, you don't want a garbage family tree because great grandma cheated on great grandad. You need a DNA match, ancestry.com does this.
You are aggressively not understanding my point.

I am not looking for a DNA accurate family genealogy, I just want an AI to remove the tedium of following sets of embedded links and recording specific data out of those links. It is something machines are quite adept, and I'd expect any half way decent machine learning to be able to follow simple instructions of where to locate the data and how to follow the links backwards. The fact that they cannot, even after being given specific instructions how means that AI is a long way off from being general use as a lot are claiming.
 
  • 1Picard
Reactions: 1 user

Mist

REEEEeyore
<Rickshaw Potatoes>
31,797
24,465
  • 1Solidarity
Reactions: 1 user

Captain Suave

Caesar si viveret, ad remum dareris.
5,605
9,632
Working 60-hour weeks will make the GPUs run faster.

Also why the fuck would people want to work 60-hour weeks to build a thing to replace themselves?

AI is nothing but the Great Replacement on steroids.

- All the worker productivity studies show people perform best at ~35 hours a week

- If it's so valuable a goal, why not hire 50% more people?

etc, etc.

"Many of you will burn out. Some may kill yourselves. Ideally, all of you will eventually be replaced. But it is a sacrifice I am willing to make."