In which Ann Leckie sums up my dissertation in about seven skeets. - ThreadSky

wolvendamien.bsky.social • 61 days ago

In which Ann Leckie sums up my dissertation in about seven skeets.

Seen a few folks astonished by my lack of support for ChatGPT and similar. Because AI, right?

Yeah, no. What we have right now is glorified autocorrect. Glorified autocorrect that takes the energy of an entire city to do what it does, plus half of Lake Erie to cool down while it's doing it.

Comments

wolvendamien.bsky.social•61 days ago

Seriously tho: Even if LLM/GPT-type "AI" was anything close to a nonbiological mind, it'd still be one built & reared by & meant to be in permanent oppressive extractivist service *of* billionaire capitalists & soon as said captialists realise "AI" isn't that, they'll just keep marginalizing humans.

wolvendamien.bsky.social•61 days ago

Here you go:

'Belief, Values, Bias, and Agency: Development of and Entanglement with "Artificial Intelligence"'
https://vtechworks.lib.vt.edu/handle/10919/111528

cdrhbiscuitiii.bsky.social•61 days ago

Also highlights another key problem. Everything is being bullshitised. They don't care that jobs are done well, just that things are done cheaply and sold with a high margin.

webrant.bsky.social•61 days ago

It's a binary pattern match masquerading as "thought" by con men ignoring the built-in regression when the model scrapes data tainted by itself

If it wasn't also a tool for short term capital concentration with the promise of subjugation we wouldn't even be talking about it
https://bsky.app/profile/wolvendamien.bsky.social/post/3leghocysj22z

hominid.bsky.social•61 days ago

Is it reasonable to think that the current project behind chatgpt is to teach the machines to manage human emotions? I “chatted” with chatgpt about a history question; it seemed much more focused on discerning and managing my emotions than on finding correct answers.

wolvendamien.bsky.social•61 days ago

I would look into sentiment manipulation experiments by groups like fb and google, and then ask yourself what openAI gains by presenting people with what feels like and emotionally responsive interlocutor

wolvendamien.bsky.social•61 days ago

*an

hominid.bsky.social•61 days ago

That makes sense.

Along those lines — I watched a French and Saunders sketch on Youtube yesterday. Then watched the two videos that Youtube queued for me — a Mary Tyler Moore scene and an SNL sketch. The only common thread — all had a woman saying she wished she were dead. Grim.

adamsith.bsky.social•61 days ago

You must have mixed feelings about that.
(At least it wasn't Chat GDP doing it)

iddqd58.bsky.social•61 days ago

Honestly, the bar to clear for us to trust any further "AI" innovations should be fixing the Outlook search function.
Get that right and maybe we'll listen.

kirke-apologist.bsky.social•61 days ago

I've literally had the same rant as a software engineer and I'm losing my mind. I'm so sick of this garbage they're cramming down our throats. I treat it on the same level as websites with too many popups, ads that force you to accidentally click on them, and other forms of obnoxious UI spam

ranndomx.bsky.social•61 days ago

ChatGPT isn’t a real AI but it’s more useful than a spell checker. I use it at least once every couple weeks to ask it about pretty complex C# coding issues and more often than not it churns out good code. It’s not going to replace software developers but it’s a lot better than stackoverflow.

kevincureton.bsky.social•61 days ago

I’d have less of an issue if the whole thing wasn’t built on copyright violations at the core. Is there a code AI that can back up the legality of their training data? Legit curious.

ranndomx.bsky.social•61 days ago

I absolutely do not approve of the copyright violations they committed in training their AIs and I have no problem with them suffering legal consequences for it. But I also can’t pretend it’s a just glorified spell checker.

mjwhitehead.bsky.social•61 days ago

I mean it might potentially save you time from stackoverflow, but it's unlikely to give you a better answer, and over time it will kill the sources of data it's used to farm said answers by being quicker, so fundamentally you're just in a good window for coding solutions right now.

ranndomx.bsky.social•61 days ago

It gives more concise answers than I find on StackOverflow and it can write some really complicated snippets of code and SQL scripts. I’ve used StackOverflow since its inception, but ChatGPT is head and shoulders above it in usefulness. It’s well beyond “farming” answers and regurgitating them.

judgewooden.bsky.social•61 days ago

Staying on chatgpt does mean less engagement on stackoverflow, and less chance to contribute. I wonder what the future is of that type of shared community?

mjwhitehead.bsky.social•61 days ago

That's all it's doing though, farming answers, averaging out content, and regurgitating it. Good for you that you find the averaging useful, but it will kill off the places where those answers are discussed if everyone uses it.

Comments

Posting Rules

Reply