Boosted by @MaryAustinBooks@mstdn.social
Replying to
every conversation about the potential usefulness of AI, divorced from ethical concerns, is just this dril tweet
Remote
Antoine Leblanc 
@nicuveo@tech.lgbt
mastodon 4.7.2+glitch.techlgbtFriendly neighbourhood queer nerd, wanted Haskell criminal.
My cutie mark is a lambda.
333 Followers
114 Following
15 Posts
Joined November 22, 2022
pronouns:
they/them
blog:
everywhere else:
location:
Dublin, Ireland
RE: @joshuaelliott@mastodon.social
this is the main reason why i believe that chatbot addiction / chatbot psychosis is a LOT more widespread than we realise: people with a clear understanding of the ethical issues try claude once, it does a thing correctly enough, they get one-shot, and they start posting like if sephiroth was on linked-in, ethical concerns be damned. it keeps happening.
it's a big part of why i refuse to touch LLMs: i am not special, i am not magically immune to the brain worms that have claimed so many smart people. if it happened to them it could happen to me.
Replying to
@balinares@furry.engineer i... don't think i'm shouting at people on our side?
i am describing a phenomenon that i've witnessed multiple times, whereby someone ostensibly on our side tries claude and then *enthusiastically switches side* and becomes an AI booster. i am therefore suspecting that there is something deeply seductive / addictive about chatbots that results in people abandoning their previously held positions. and that is one of the reasons why i avoid using chatbots, because i am afraid it might work on me.
i do get your point about not being completely out of touch to be more convincing, but... you don't need to literally be a recovering gambling addict to reliably warn people against gambling addiction, you just need to be informed enough, you know? and you can get informed enough without spending every night at the casino.
There was something that bothered me about mastodon, coming from twitter, but that i was not really managing to articulate until it hit me this morning: mastodon is profoundly ephemeral.
Moving across servers is common; and when you do all of your post history remains on the old server and is probably gone. Most servers disable text search. Between the general lack of discoverability, the lack of lists, and the unaltered chronological timeline, toots from friends can just simply be lost in the flood, never to be seen.
As a result it also doesn't work as a "repository" of information, it could never be a place where dril does his thing, for instance: how are you going to find the old dril toot you want to drop in a conversation? Scroll through his history?
The only thing that matters is your transient feed. And it's not a bad thing per se! I think most of this is on purpose. This is also how a lot of people used twitter... but not how i did. And i have to get used to that.
Replying to
@jenniferplusplus@hachyderm.io yep, it's been down for a couple hours now; their status page claims that all systems are operational, but acknowledges that some error rates are high. but in practice i can't even log in.
Replying to
@typeswitch i am sorry :(
it really is infuriating to see european countries follow into the moral panic one by one...
Replying to
@balinares@furry.engineer i guess you could read my point about why i avoid the use of chatbots altogether as an admonition; but FWIW that's not what i intended? it's not a moral judgement, more of a warning: if you choose to / have to use them, be careful, we don't understand AI psychosis yet but it's a real thing.
Replying to on hachyderm.io
@vie@hachyderm.io i wonder why that's the case, but yeah i've observed that too. my gut feeling is that code is extremely predictable, so LLMs are better at that than they are at other things, making them more enticing? and then the chatbot psychosis sets in.
Replying to
@LionsPhil@plush.city i disagree. for instance: if i see a screenshot of a tweet making the rounds, i can go look for that tweet. if the account has changed handles, it can be a bit of a challenge, but i'll be more often than not able to find the original tweet, and link it, no matter how old. on here, without text search, that's basically impossible. if the instance still exists i can find the account, but i'm not gonna scroll through three years-worth of toots to find the original one i'm looking for.
Mastodon cannot have a "milkshake duck" toot, or a "leopards eat my face" toot, or replicate the success of dril.
Replying to
@maegul@hachyderm.io i think twitter already was in this weird middle-ground; it was never intended to become a repository of information, but it became one nonetheless... but, otherwise, i agree: mastodon isn't a "public square". it's more private. which, honestly, is pretty great? but that just means it can't be a 1:1 replacement for twitter: influencers, politicians, and people who want to build their brand will never get the same reach here.
Replying to
@balinares@furry.engineer @LionsPhil@plush.city you troll :D
but, seriously, what i failed to properly articulate, is that i don't think any of this is a bad thing! no text search helps prevent harassment, chronological timeline is what i want! it's just that i have to get used to the fact that this place is different, in mostly good ways. ^^
Replying to
@pmorinerie @jcoglan
so, two things on this. On one hand: i agree! arguing against genAI / LLMs in terms of efficiency / productivity is arguing *on their terms*. it's implicitly accepting that those tools are okay to use, despite the fact that even if they were good there would still be a mountain of objections.
on the other, i think you've missed the original point that OP was making. :D
i don't read the original point as "LLMs are a bad layer of abstraction because they generate bad / bloated code"; i read it as "by their very nature, they are *random* processes: you cannot treat them as just another layer of abstraction, because *every output has to be reviewed*". no matter how bad a compiler is, it is predictable: sure, it will generate inefficient ASM, but it will always generate ASM the same way, while a LLM *by design* cannot.
Replying to
@1password@1password.social hint: the ideal "agent to human ratio" is 0. seeing a company jeopardise its reputation like this just to chase a trend is very disappointing.