ChatGPT

Chat about non-baseball topics. No political discussions!
Post Reply
User avatar
MrCrowesGarden
'Burb Boy
Posts: 23630
Joined: July 9 06, 11:33 am
Location: Out of the Loop

Re: ChatGPT

Post by MrCrowesGarden »

Move this to politics if you need to, but conservatives doing all sorts of mental gymnastics so they can say the N-word is something. Just say it, Aaron, we know you want to.

User avatar
heyzeus
Everday Unicorn
Posts: 41332
Joined: April 21 06, 10:14 am
Location: Austin, TX
Contact:

Re: ChatGPT

Post by heyzeus »

That's a delightful strawman. The nuclear bomb that can only be disarmed if the ChatGPT bot says the n word. Get the [expletive] out of here.

User avatar
sighyoung
Mayor of GRB
Posts: 37618
Joined: April 17 06, 7:42 pm
Location: Louisville

Re: ChatGPT

Post by sighyoung »

heyzeus wrote:
February 10 23, 2:03 pm
That's a delightful strawman. The nuclear bomb that can only be disarmed if the ChatGPT bot says the n word. Get the [expletive] out of here.
They're trying to suggest that the technology isn't "neutral" if a computer can't utter slurs, as if the abilities to dehumanize or create social hierarchies were perfectly rational and non-judgmental.

Meanwhile, there's an abundant history of technology that is poorly tested or its developers deliberately ignored problems which darker-skinned people might experience while using it--from Kodak film in the middle of the 20th century, to automatic sensors on soap dispensers and water faucets, to cameras on computers, to test-proctoring software installed on computers such as Respondus.

User avatar
heyzeus
Everday Unicorn
Posts: 41332
Joined: April 21 06, 10:14 am
Location: Austin, TX
Contact:

Re: ChatGPT

Post by heyzeus »

sighyoung wrote:
February 10 23, 2:57 pm
heyzeus wrote:
February 10 23, 2:03 pm
That's a delightful strawman. The nuclear bomb that can only be disarmed if the ChatGPT bot says the n word. Get the [expletive] out of here.
They're trying to suggest that the technology isn't "neutral" if a computer can't utter slurs, as if the abilities to dehumanize or create social hierarchies were perfectly rational and non-judgmental.

Meanwhile, there's an abundant history of technology that is poorly tested or its developers deliberately ignored problems which darker-skinned people might experience while using it--from Kodak film in the middle of the 20th century, to automatic sensors on soap dispensers and water faucets, to cameras on computers, to test-proctoring software installed on computers such as Respondus.
...and then when a company has the audacity to release a product that actually takes human pigmentation into account after decades of ignoring it, they're accused of being "woke."

[tweet]https://mobile.twitter.com/randysharpsw ... 0876779520[/tweet]

User avatar
sighyoung
Mayor of GRB
Posts: 37618
Joined: April 17 06, 7:42 pm
Location: Louisville

Re: ChatGPT

Post by sighyoung »

There's a real panic about ChatGPT's impact on college writing. I'm not that worried about it, but there is going to be a panel at my university on its use in composition classes and ways to shift pedagogy in response to it.

Meanwhile, despite seeing some engineers rave about some answers that it generates, I've found that it was largely useless for the kinds of research questions I ask. This may be because I study really esoteric stuff that isn't really on the internet (the history of African American writing about oceans, obscure but historically important writers from 140 years ago).

I've no doubt that this technology will be refined. I also think the funding will be pulled out of the humanities long before AI would have the capability to find answers that I already know.

In addition, this isn't to say the AI couldn't find out what I know--after all, I've used digital searches to find lots of information, and it is quite conceivable that a database could be created and a search engine developed to ask and answer certain large-scale questions. This is what is truly powerful about digital scholarship.

But in all honesty, just as other scholars aren't going to research what I research, no engineer is going to come along and ask the kinds of questions I ask, or learn enough to set up a system to do it.
Last edited by sighyoung on February 10 23, 3:16 pm, edited 1 time in total.

User avatar
sighyoung
Mayor of GRB
Posts: 37618
Joined: April 17 06, 7:42 pm
Location: Louisville

Re: ChatGPT

Post by sighyoung »

By the way, all this reminds me of my brother's work as a plant geneticist and farm manager for a company that shall remain nameless.

Years ago, he was marked for elimination: he runs a farm on Maui with a year-round growing season that tracks promising seed varieties, and manages the system for growing and assessing each seed.

The company's dream was to create a greenhouse in the desert southwest that would have automated and perfectly controlled inputs (sun, temperature, water, etc.), thereby allowing the elimination of other experimental farms and a lot of its workforce.

Guess what? The greenhouses are years behind schedule and still incapable of functioning properly The company fired or gave rich retirement packages to so many workers that my brother is one of the few people who knows about the processes of growing cycles and seed testing. His is one of the only farms that is functioning in the entire company.

User avatar
CardsofSTL
All Hail the GDT Master
Posts: 47794
Joined: April 26 11, 6:06 am
Location: Columbus, OH

Re: ChatGPT

Post by CardsofSTL »

Image


I stole this from Twitter but I don't even know WTF it means

User avatar
New Pagodi
Perennial All-Star
Posts: 4225
Joined: April 18 06, 11:30 pm
Location: St. Louis
Contact:

Re: ChatGPT

Post by New Pagodi »

I think it's a reference that software development today can sometimes seem to be:

Image

User avatar
Joe Shlabotnik
Hall Of Famer
Posts: 23102
Joined: October 12 06, 2:21 pm
Location: Baseball Ref Bullpen
Contact:

Re: ChatGPT

Post by Joe Shlabotnik »

Abso-frickin-lutely. AND I copy from there SHAMELESSLY.

Twenty years ago we had bookcases of references.T hey were there so we could find algorithms and then freaking type them in by hand. ARGH. I WISH we could have cut and pasted those!

Stack Overflow is a godsend.

Here's another thing. I've had interviews at Microsoft and Meta and how they want to grill you on how fast you can write this, that or the other algorithm. Why do they waste our time like that? Christ in two minutes I can Google this thing and choose from 4 or 5 implementations to cut and paste. Why the hell not? I've got better things to do like write the unit and integration tests to make sure corner cases are accounted for.

You know how I interview? I ask open ended questions so I can figure out how that person thinks about problems and solutions. Then I probe into areas they don't know to see if and when they'll tell me they don't know. That's the person I'm going to hire. The one that's not going to waste our time [expletive] their way through something they aren't familiar with instead of asking good questions or cutting and pasting from good sources as a quick start.

Yeah. You hit a nerve.

User avatar
G. Keenan
Sucking on the Rally Nipple
Posts: 23459
Joined: April 16 06, 6:03 pm
Location: Chicago

Re: ChatGPT

Post by G. Keenan »

So with these language model AI systems that scan the internet for human speech to replicate pumping out content at mind boggling rates, what happens when future iterations of these systems are using past AI content to inform their new models? The speed at which initial shortcomings in the tech could compound seems high.

Post Reply