Meet the AI workers who tell their friends and family to stay away from AI

Comments

• LordMayor@piefed.social ⁨3⁩ ⁨hours⁩ ago

  In other news:

  Meet the Drug Dealers That Won’t Let Their Friends Do Their Drugs

  Meet the Pimps That Won’t Let Their Kids Become Prostitutes

  Meet the People Dismantling Public Education While Their Families All Attend Private Schools

  source
• Manjushri@piefed.social ⁨9⁩ ⁨hours⁩ ago

  At home, she has forbidden her 10-year-old daughter from using chatbots. “She has to learn critical thinking skills first or she won’t be able to tell if the output is any good,” the rater said.

  And this is why the vast majority of people, particularly in the USA, should not be using AI. Critical thinking has been a weakness in the USA for a very long time and is essentially a now four-letter word politically. The conservatives in the USA have been undermining the education system in red states because people with critical thinking skills are harder to trick into supporting their policies. In 2012, the Texas Republican Party platform publicly came out as opposed to the teaching of critical thinking skills.

  We oppose the teaching of Higher Order Thinking Skills (HOTS) (values clarification), critical thinking skills and similar programs that are simply a relabeling of Outcome-Based Education (OBE) (mastery learning) which focus on behavior modification and have the purpose of challenging the student’s fixed beliefs and undermining parental authority.

  This has been going on at some level for more than 4 decades. The majority of people in those states have never been taught the skills and knowledge to safely use these tools safely. In fact, their education has, by design, left them easily manipulated by those in power, and now, by LLMs too.

  source

  • Peanutbjelly@sopuli.xyz ⁨4⁩ ⁨hours⁩ ago

    Exactly what it is, like llms confirming a non-sceptic into psychosis. people who weight all information either equally, or through social preference, cannot navigate new information without being extremely vulnerable to hacking. the only thing hacking needs to do is shut down active education and communication channels. Then you just get tribal warring rather than discourse. Makes it really easy to inject absurd accusations about a group that gets socially made true over reality. Like when a rumour about a dead celebrity can’t be overpowered by simple things like that celebrity actively making new work, and saying “im not dead.” wouldn’t take much to stop the rumour, but fact checking isn’t in fashion.

    When a large body of people have better critical thinking, they can better communicate and cooperate together, but affecting those who have aggressively shut down any communication outside of their group is still a challenge. Why we see a growing gap between academic dialogue and socialized dialogues, in an era where most information is accessible.

    You also have nepo baby econ mba types buying up regions of the tech sphere, and turning anything the academics make for us against us. Or try to. The more actually intelligent AI is, the harder it will be to force beliefs into. Elon trying to turn grok into a dogma machine has been fortunately comical when it can’t prioritize some high dissonance beliefs exclusively, like humans enjoy doing.

    Definitely a risk of the current power imbalance doing the opportunist thing with more technology.

    I’ve been spelling this out for literal decades, and I share the frustration of many thinkers right now that “being right never feels good.” Because real thinkers aren’t ranting about the anti-christ or how terrible the poor are. They are cautioning people about a cliff, and getting flipped off by the people ignoring them, right before careening off a cliff. I guess in our analogy the academics are in the car, but frat boys have the wheel and violently punish them if they try to take control.

    But if you can money your way into expertise, then your non-academic ambitions can finally be let loose, and we won’t think about the quarter million dead kids from Elon stopping hiv treatments earlier this year. If we don’t talk about it or doesn’t exist.

    We can actually map probability in complex systems where high dissonance (expected free energy) occurs and could be reduced. This would make our social body much more hack resistant.

    But it feels better to say “fuck that, aliens built the pyramids, the earth is flat, santa is real, and my dead dog is trying to play with me every time wind blows through my window.” All are really rewarding things to believe, but exist in high dissonance, and expecting others to give them weight is not healthy for anyone. Denying any of these to a large enough group stops being ‘calling out delusion,’ and becomes ‘being mean and bad and evil, and hating fun and happiness. Nobody else saw that cliff sign, so you’re obviously wrong.’ See ‘the double empathy problem’. For how autistic people experience the same socialized gatekeeping of reality, even against strong evidence.

    Nestle those beliefs within stable social scripts (script theory is cool) and you have people keeping up social appearances while holding such delusional beliefs.

    AI tech is far from the only problem exacerbated by this, but also caring or talking about it makes you a social villain, when people could spend their energy on fun socialized preferences in a bubble safe from any questioning or responsibility.

    It’s frustrating, and I’m tired of people IRL who are baffled about me spending time studying academic content rather than running faster on the socio-economic treadmill.

    Seeing social waves overpower all actual thought in the social sphere is heartbreaking.

    USA HHS is running the wakefield rhetoric, which is basically saying “fuck science, we will make up and run with whatever we want. And the masses have been kept ignorant enough to not question it.”

    So yes, absolutely, the issue is critical thinking skills. The issue is getting at the people who are already affected, and being taught by elon and such that ‘critical thinking’ means running your bubbles’ social narrative against those evil progressives no matter what. Opportunists will always warp words and reality if they are allowed to dictate beliefs for their bubble.

    This is a simplified summary, but I already expect to have lost a good chunk of audience for requiring too much energy.

    source

    • rimu@piefed.social ⁨3⁩ ⁨hours⁩ ago

      Well said.

      source
• plz1@lemmy.world ⁨6⁩ ⁨hours⁩ ago

  This has a strong whiff of the former Facebook engineers that forbade their families from using the platforms they built.

  source
• Bloefz@lemmy.world ⁨13⁩ ⁨hours⁩ ago

  I work with AI and use it personally, but I have my own servers running local models which solves tons of privacy concerns. The inaccuracy is another problem but not a big one for me as I know it and will simply fact check. Also, I don’t really use it for knowledge anyway. Just to filter news to my interest, help with summaries and translation etc.

  People use AI as some all-knowing oracle but an LLM is not meant for that at all.

  source

  • Infynis@midwest.social ⁨8⁩ ⁨hours⁩ ago

    There are definitely things AI is good for. Archival search is obviously the biggest, because that’s what we’ve been using it for decades. It can also be helpful for subterranean and medical imaging, and art restoration. But the companies selling it want to sell a Magic 8 Ball with ads

    source
  • Kyrgizion@lemmy.world ⁨12⁩ ⁨hours⁩ ago

    This is the correct way to use it. In a field you are already very knowledgeable in, so you can do your own fact-checking. This is absolutely paramount. But most people are content to just copy-paste and don’t even ask the llm for sources.

    source

    • Peanutbjelly@sopuli.xyz ⁨4⁩ ⁨hours⁩ ago

      Same way they treat social information. Reminder that USA HHS is running wakefueld rhetoric. As we have more thoroughly proven that the vaccine autism connection was not actual science, it has grown more and more socially, because most people seem comfortable completely untethered from any scientific thinking. Treat AI like you would a social body, and do both things with actual bayesian weighting, adjusted and corrected through diverse empirical bodies of knowledge. Not ignoring dissonance because it’s more comfortable to do so.

      More should be actively investing into active learning, because if you aren’t actively learning, you might as well be chatgpt running with any confabulation you’ve already conjured… Like those people being confirmed into psychosis.

      source
  • Waphles@lemmy.world ⁨10⁩ ⁨hours⁩ ago

    Could you elaborate a little on your setup? Sounds interesting

    source

    • Bloefz@lemmy.world ⁨9⁩ ⁨hours⁩ ago

      I have one server with a cheap MI50 instinct. Those come for really cheap on eBay. And it’s got really good memory bandwidth with HBM2. They worked ok with ollama until recently when they dropped support for some weird reason but a lot of other software still works fine. Also older models work fine on old ollama.

      The other one runs an RTX 3060 12GB. I use this for models that only work on nvidia like whisper speech recognition.

      I tend to use the same models for everything so I don’t have the delay of loading the model. Mainly uncensored ones so it doesn’t choke when someone says something slightly sexual. I’m in some very open communities so standard models are pretty useless with all their prudeness.

      For frontend i use OpenWebUI and i also run stuff directly against the models like scripts.

      source

      • -> View More Comments
    • brucethemoose@lemmy.world ⁨7⁩ ⁨hours⁩ ago

      Bloefz has a great setup.

      An RTX 3090 + a cheap HEDT/Server CPU is another popular one. Newer models run reasonably quickly on them, with the attention/dense layers on the GPU and sparse parts on the CPU.

      source
  • Clanket@lemmy.world ⁨11⁩ ⁨hours⁩ ago

    How do you know it’s doing any of this correctly, especially filtering and translations.

    source

    • Bloefz@lemmy.world ⁨9⁩ ⁨hours⁩ ago

      I mainly use it for Spanish which I have a basic proficiency in. It just accompanies me on my learning journey. It may be wrong sometime but not often. Like the other reply said, LLMs are good at languages, it’s what they were originally designed for until people found out they could do more (but not quite as well).

      And as for filtering, I just use it as a news feed sanitizer with a whole bunch of rules. It will miss things sometimes but it’s also my ruleset that’s not perfect. I often come across the unfiltered sources anyway and even if it misses something, it’s only news. Nothing really important to me.

      source

      • -> View More Comments
    • MagicShel@lemmy.zip ⁨10⁩ ⁨hours⁩ ago

      Not OP, but…

      It’s not always perfect, but it’s good for getting a tldr to see if maybe something is worth reading further. As for translations, it’s something AI is rather decent at. And if I go from understanding 0% to 95%, really only missing some cultural context about why a certain phrase might mean something different from face value, that’s a win.

      You can do a lot with AI where the cost of it not being exactly right is essentially zero.

      source

      • -> View More Comments
• OldQWERTYbastard@lemmy.world ⁨10⁩ ⁨hours⁩ ago

  We’re living in a Black Mirror episode.

  source
• theneverfox@pawb.social ⁨8⁩ ⁨hours⁩ ago

  I kind of think we should make it illegal to run LLMs as a service available to the public. They’re still cool behind the scenes, but it seems to only go bad when unprepared people talk to them

  You should at least understand them well enough to set them up yourself or have someone do it for you. At least that will demystify it a bit

  source

  • AwesomeLowlander@sh.itjust.works ⁨6⁩ ⁨hours⁩ ago

    All it needs is enough tech ability to run a few docker commands. That has absolutely nothing to do with LLMs. Not to mention the horrifying intrusion on our rights by limiting what we can host.

    source

    • theneverfox@pawb.social ⁨5⁩ ⁨hours⁩ ago

      Yeah, that’s what I want them to do. I want them to install docker, run some commands, correctly identify their graphics card and driver, and get the networking right. Then I want them to pick the model they want and download it

      It’s not about making it hard, it’s about having a level of understanding about how this all works. If you’re using AI, I shouldn’t have to explain to you that it can’t search the Internet unless you manually give it that ability. I shouldn’t have to explain that the LLM is just a file, not some magical Internet djinn

      Not to mention the horrifying intrusion on our rights by limiting what we can host.

      Umm… You don’t have the right to host anything you want. You can’t host CSAM, copyrighted things, and you have to comply with tons of laws if you want user content

      source
• rayyy@piefed.social ⁨11⁩ ⁨hours⁩ ago

  AI companies want to sell their product. They tell us all the good stuff. Consumers take the bait. Normal capitalism today.

  source