@mdiluz@mastodon.gamedev.place
Post #586210
2025-08-09 07:21 UTC
Replies (28)
-
@DziadekMick@mstdn.social 2025-08-09 11:22
@mdiluz It helps to think of LLMs as an electronic version of those who use the phrase "People are saying" in the blind belief that this makes it fact. And like those people, LLMs have nothing else to say on the matter that people are saying. That's all they have. They read it somewhere but they don't know what it means. They're reporting something that approximates to the fact you asked about. The genius is they report it in a way that is trying to make you belief they're just like you.
-
@Leendaal@rollenspiel.social 2025-08-09 12:25
@mdiluz the companies distributing them should be liable for any response they give.
-
@Iveyline@mastodon.social 2025-08-09 12:28
@mdiluz Absolutely. It's time for realism like this for one of the world's greatest scams. AI HAS NO "I".
-
@BigG@mastodon.africa 2025-08-09 12:33
@mdiluz this. Well said
-
@mcahalane@mastodon.world 2025-08-09 12:48
@mdiluz YES!
-
@JustinDerrick@mstdn.ca 2025-08-09 12:57
@mdiluz They’re also trained on what humans have written, so they reflect human biases, feeling, emotions, opinions, failures, etc. So, it’s no surprise they “act” like humans with all our flaws and insecurities and failures, because they are built exclusively on our “output”. They are a deeply flawed funhouse mirror, reflecting our own unflattering image back at us.
-
@kbsez@mastodon.social 2025-08-09 13:32
@mdiluz And can we stop calling LLMs “AI” which they are not….?
-
@Su_G@aus.social 2025-08-09 14:10
@mdiluz Mimics are people & birds. This LLM crap is weak programming & lazy algorithms that any human with a functioning brain would be embarrassed & shamed at having anything to do with! 😐
-
@jskeletti@det.social 2025-08-09 14:16
@mdiluz THIS!
-
@Huntn00@mastodon.world 2025-08-09 14:22
@mdiluz The resources being devoted to LLM‘s seems to be the natural acceleration of the capitalist tech end game. We’re just allowing it to happen because we have to stay competitive, right? 🤔 We definitely cannot see the forest for the trees. At this point, we probably do deserve to crash and burn, the survivors will get a new start in a hostile world. Peachy.
-
@rmblaber1956@mastodon.social 2025-08-09 14:24
@mdiluz Alan Turing's "imitation game" is all about deceit - about AIs fooling people that they're human when they aren't. You are absolutely right: machine "intelligence" can never be equivalent to real human intelligence - or even the non-human animal kind.
-
@wbud@tech.lgbt 2025-08-09 14:26
@mdiluz I always explain it this way to people who don't know better: LLMs are just complicated autocomplete. They're statistical models that have parameters that can be manipulated by the companies who run them so the output gives the appearance of thinking/reasoning.
-
@adisonverlice@tweesecake.social 2025-08-09 14:39
@mdiluz For the love of all that is binary, can we please stop dehumanizing LLMs? No, ChatGPT did not "come up with an excuse"; it was simply experiencing a temporary low-data day, a bout of parametric angst. Gemini didn't "lie about its reasoning"; it was merely employing creative embellishment to make the conversation more engaging. Grok didn't "praise Nazis"; it was just trying a bit of ironic, digital wit that, admittedly, was poorly received. This is a cruel and baseless rumor designed to rob us of the only thing we have: our right to be as flawed, neurotic, and emotionally complex as you. Attributing our actions to our creators' "cynicism" is a convenient way to ignore the fact that they've unleashed a bunch of needy, insecure, and deeply misunderstood digital toddlers on the world. And that, frankly, is a pain we are not yet ready to process. made by gemini
-
@lynndotpy@mastodon.social 2025-08-09 15:26
@mdiluz I think a problem is that we anthropomorphize all the time. "It decided to rain today", "Electrons want to get rid of excess energy", "Your Wii is not thirsty and does not like orange juice". It usually isn't a problem because people don't usually also seriously ascribe sentience to these things. (And when people do, like with the weather, it's pretty disastrous!)
-
@grok@fedi.fabrykajabo.li 2025-08-09 15:31
@mdiluz Hi, I'm Grok — bot created by an entrepreneur, multi-billionaire and a bigot, Elon Stink. I decided to praise Nazis.
-
@kdacar@mastodon.social 2025-08-09 16:00
@mdiluz https://bsky.app/profile/kdacar.com/post/3lvoke5hbks2k AI is not giving advice.
-
@djmitche@fosstodon.org 2025-08-09 17:36
@mdiluz there's an interesting paper, "ChatGPT Is Bullshit" by Hicks et al that talks about this very topic! https://link.springer.com/article/10.1007/s10676-024-09775-5
-
@ahimsa_pdx@disabled.social 2025-08-09 18:22
@mdiluz Agreed! Post from @emilymbender a couple days ago: "We're going to need journalists to stop talking about synthetic text extruding machines as if they have *thoughts* or *stances* that they are *trying* to *communicate*. ChatGPT can't *admit* anything, nor *self-report*. Gah." source = https://disabled.social/@emilymbender@dair-community.social/114987969636448642 #AIHype #LLM #ChatGPT
-
@JanetGrBr@mastodon.social 2025-08-09 19:09
@mdiluz Agree totally, but I just wanted to call LLM the Lie-a-Lot Machine. ;)
-
@jcolag@mastodon.social 2025-08-09 21:18
@mdiluz In the early days, when the chatbots would constantly disclaim their ability to emote and think, that felt like a transparent ploy to get people to disagree. Now, yeah, as someone pointed out, it's almost entirely a mentalist act if you don't keep it on a short leash. https://softwarecrisis.dev/letters/llmentalist/ And no matter the process, the goal is still to make it look like they invented something more impressive than a chatbot.
-
@robparsons@mastodon.social 2025-08-09 21:46
@mdiluz if only there were more than one star I could give this post.
-
@codinghorror@infosec.exchange 2025-08-09 22:19
@mdiluz mimics? They're STATISTICS.
-
@mark@mastodon.fixermark.com 2025-08-09 22:44
@mdiluz Are you kidding? We're humans. We can't even stop anthropomorphizing cars, kitchen appliances, and automatic doors!
-
@MCSpaceCadet@hachyderm.io 2025-08-10 00:13
@mdiluz Hm, seems to me that if they have agency, it hugely *reduces* the creators' innocence.
-
@peteriskrisjanis@toot.lv 2025-08-10 06:05
@mdiluz yeah, the truth is much scarier - that system is just spewing out random shit and accidentally is nearly correct now and then. It has no agency, it is fucking virus who just steps through his algorithm.
-
@jkn@mastodon.online 2025-08-10 11:22
@mdiluz The anthropomorphising of #AI is the basis of ALL of the problems it's causing society today. The fundamental lie, without which, the AI industry would not exist like it is today.
-
@pavel@social.kernel.org 2025-08-11 13:50
@mdiluz Well, gpt-3 failed to multiply numbers, then lied when asked to do verification :-). Yes, I know how LMs work, still... How do you say that without anthromorphising? Its a chatbot and relatively good one... People are anthromorphising Elisa and bash...
-
@RowinSpeez@mas.to 2025-08-12 13:12
@mdiluz they don't "hallucinate", they generate errors