all 10 comments

sorted by: hot top controversial new old
[–] 16 points 1 day ago (1 child)

"We investigated with urgency and determined that this was caused by a bug introduced in the product that caused the model to hallucinate and was NOT a security or privacy issue," Singleton wrote.

Yeah I believe that. They fucked something up and the random text generator went off generating some random text. Who could have seen this coming?

Additionally, no one should be trusting a piece of software that's basically the linguistic equivalent of a Roomba trying to navigate a living room with any sort of action capability. As enough people use it, eventually one time it will say "yes, deleting all your emails is a good idea".

  • source
  • hideshow 1 child comment
  • [–] 5 points 1 day ago

    This is a case where actually the LLM was correct, deleting all your emails is absolutely the correct thing to do. Even better, it gives you a valid excuse if anyone ever asks. You gotta stop making the case for AI so easy, man

  • source
  • parent
  • [–] 10 points 1 day ago

    Working as designed.

  • source
  • [–] 9 points 1 day ago (1 child)

    Reuters worked with staff at Signal and the AI Now Institute to test if they could replicate that Meta staffer’s and others’ Muse experiences. On Android devices, they found that Muse could read Signal messages by bypassing Meta’s “Connector” tool and instead relying on permissions granted during initial setup.

    On iPhone, requests were blocked, Reuters reported. But on Apple’s macOS, Muse repurposed “settings Apple offers to help people with disabilities operate their computers” to retrieve messages, including screenshot tools.

    Yikes

  • source
  • hideshow 1 child comment
  • [–] 8 points 1 day ago (1 child)
  • [–] 12 points 1 day ago

    Metas chat box, your ‘ai personal assistant’ or otherwise the gaping privacy nightmare that might post your credit card on X, the everything app, or send your intimate photos to your weird uncle on Facebook, or any number of other miserable things.

  • source
  • parent
  • load more comments (-1 replies)
    [–] 6 points 1 day ago

    ... I should ask Grok what Elon's bank account numbers are.

  • source
  • [–] 4 points 19 hours ago

    Look, I know this is corporate negligence, but look how friendly and fuzzy that guy is!

    I’m practically jamming my credit card and tax forms at the screen whenever he appears.

  • source
  • 20 comments in c/technology@lemmy.world view that post
    [–] 35 points 21 hours ago (11 children)

    I really hate that this article is pretending that Muse is somehow "spying" on users.

    This isn't spying. This is just bad engineering.

    An LLM should only be given access to relevant information for a particular query. If you give an LLM access to irrelevant information, it gets confused. Anyone who's used LLMs much knows this. It's kind of baffling to me that the project managers over at Meta don't understand how dumb their design is.

  • source
  • hideshow 11 child comments
  • [–] 32 points 21 hours ago* (4 children)

    From experience, chances are that this is indeed spying, and not just "bad" engineering.
    Meta has been about spying from day 1. With Zuckerberg laughing his ass off, that people freely give away sensitive and personal information.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 7 points 21 hours ago (3 children)

    I'm not saying Meta isn't spying on users. I'm sure they are. But this doesn't make sense as a symptom of spying.

    If you want to spy on users you just record all their info. You absolutely don't give the LLM they talk to access to that same information - that's nearly useless, and is one of the surest ways to make sure the users find out. The LLM doesn't need to know about the spying.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 8 points 21 hours ago (1 child)

    You're assuming they're not dumb af.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 7 points 18 hours ago

    It's both, and they are indeed dumb AF. Muse drinks from the same spying trough the rest of the slop platform sits on.

    Where they're being idiots is that the model is internally being fed a system prompt meant to foster "engagement" in between "sessions", and it absolutely is being fed not just context beyond the scope of the task, but that context...

    Is clearly being populated by another LLM model that is slurping from behind where the totally secure e2ee decode happens after the "private" message goes from Grandma Conspiracy Hoax's inbox and through Aunt Zionist Shill onto Zuckerberg's private masturbation window.

    The second model doesn't match the info against the specific user before passing it to the first, it passes the full search results it pulled and leaves the first model to determine relevant use.

    Great shit if you're running a tight local harnessing set up and don't want to waste compute on your helpers sanitizing the a2a calls that the model handling orchestration would need to parse anyways.

    Asinine and dumbfucked when your set up is a multi-user activity panopticon which your bottom line desperately relies on pretending isn't the case.

  • source
  • parent
  • [–] 6 points 17 hours ago

    We know that some AI companies are trying to use it to build a system to profile users for, e.g., monitoring protesters, and now that one of the most privacy invasive, data hungry tech companies got invasive again, you believe it was just a whoopsie doodles? Because I do not.

  • source
  • parent
  • [–] 5 points 14 hours ago

    Wait until you read all of the lawsuits Facebook has lost and you will realize everything they do is only for the intent of illegally spying. They try soooo hard to go out of the way to collect and log as much data on you as possible. Many of which is illegal or should be if its not.

  • source
  • parent
  • [–] 3 points 11 hours ago* (1 child)

    It's Meta, so spying is to be expected.

    But it also seems a pretty negligent instance of engineering.

    Signal President Meredith Whittaker told Reuters that Meta launched Muse “without showing any respect for basic privacy or cybersecurity.”

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 7 hours ago

    In the case of at least some of these AI's, I expect that private data (especially messaging etc to train them to seem more"human") was probably used in training, making it pretty much built-in. That's likely as much a management decision as an engineering one.

  • source
  • parent
  • [–] 6 points 20 hours ago (4 children)

    so we now have the first AI with dementia. that is... frightening.

  • source
  • hideshow 4 child comments
  • [–] 21 points 20 hours ago (3 children)

    AI = Weighted text generator

    There isn’t anything there to HAVE dementia. We must remember this is random token generation, and not an intelligence as we otherwise know it

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 12 hours ago (2 children)

    Bro, most people think their dog understands them. AI’s goal isn’t to achieve sentience, it’s convince the average dumbass they they can extend that conversation with their dog to their chatbot.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 1 point 12 hours ago (1 child)

    Dogs do understand people they're literally evolved to do so. LLMs don't have a mind at all and are only capable of outputting and inputting text, and any emergent properties.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 10 hours ago

    I think you’re being quite pedantic. You know what I mean, dogs do not understand the grammer rules of languages, they cannot carry on a conversation without someone “interpreting” the dog’s response. Obviously I get that dogs know commands, jesus.

  • source
  • parent
  • [–] 3 points 11 hours ago

    Yup.

    And they knew about these sorts of problems BEFORE Muse was released. Made a video about it a few weeks back.

  • source
  • [–] 4 points 15 hours ago (1 child)

    Why do people keep working for these asshole companies? I believe if you work for them you got terrible values, or none at all

  • source
  • hideshow 1 child comment
  • 3 comments in c/fuck_ai@lemmy.world view that post
    [–] 14 points 20 hours ago

    For reference, these staffers were already comfortably up inside the asshole of anyone who uses Facebook shit. So this must be pretty bad

  • source
  • [–] 8 points 19 hours ago

    So do something about it. Workers have a lot of power if they'd use it.

  • source
  • [–] 2 points 15 hours ago

    This doesn't read like an ad, this reads like zuck gave it a primary directive to succ up any messages it could while lying about it. Wild stuff.

  • source