Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)A
Posts
6
Comments
647
Joined
3 yr. ago

  • Both produced in South Korea? 🤷‍♂️

  • First of all, I'll own my bad - I used the term "fine-tune" in a general sense. I didn't mean to muddy the waters and I wasn't referring to the fine-tuning stage of the neural network.

    You're right about it being a cheaper fix than retraining the model, with the duct tape boat analogy - this is exactly what I've been saying. The goblin lines have been added to address a specific issue that was noticed with the latest release - it's a stop-gap.

    And yes I've seen the full list of background instructions - the first thing I did after reading the article was to check on GitHub to confirm that it's true because it sounded so bizarre.

    There isn't a huge list of instructions of topics or shouldn't cover. There are a lot of instructions about how the agent should behave but there is not a massive list of keywords / topics to avoid as you're claiming.

  • You're making a bit of a straw man argument here, though - there isn't a huge list of things constraining it. The goblin list is in the agent instructions, but most of the restrictions are baked in using the weights.

    The goblins etc were added to the list to address a specific problem. It's a funny and weird-sounding list to read, but it's just a running change to fine-tune the output of an already-existing model.

  • By "made out of tissue paper", I assume you mean written in a list in English?

    These lines were added to the agent instructions to address a specific weird behaviour that had been observed in Codex's output. How would you have done it correctly?

    Filter the output to remove all instances of raccoons? What if the project is actually about racoons?

    Run an adversarial LLM specifically to double check and, if necessary, correct instances of racoons? Using twice the power and still needs to be defined in text.

    Train a new model with an anti-racoon bias? I'd be surprised if they didn't for the next iteration, but it takes time.

    The reality is that for something this daft, the immediate fix is this.

    Biases against outputs that might encourage self-harm, murder, etc are baked into the models during training nowadays. These guardrails are there in the neural network, not as text or instructions, but part of the structure itself.

    The plain text agent instructions just give the different models a push in the direction that they want. Apparently it was mentioning racoons in unexpected contexts, so for now they just told it not to anymore.

  • No, controlling the behavior by providing a hand-tuned list of no-nos shows that we have no idea how to make an AI stay on task.

    This is how you make an AI stay on task. This is how literally anything with any semblance of uncertainty is programmed - you provide guide rails to keep the behaviour confined to whatever you want to limit it to. It's just we don't normally used plain English to do it.

    Even koopa trooper in super Mario Bros has guide rails dictating the limits of its behaviour. They turn when they reach blocks. Red ones turn when they reach a ledge. That's for something that literally just moves across a screen horizontally. ChatGPT is trained to output text for a nearly unlimited number of topics - it's insane to think that you wouldn't need a fair number of guide rails.

  • nice

    Jump
  • The idea of homework is to condition us to accept poor work-life balance.

    By all means make stuff available for students who want it, but if students need extra work that can't fit in class, then the school day should be longer.

  • This was perhaps the most surprising thing for me too - just how little progress players make in each session compared to what I expected.

    Also I was surprised by their interactions with NPCs. The first goblin they captured was made to lead them to his lair, whereupon his throat was slit. The second goblin they captured was made to march about in front of them as a trap detector before they eventually adopted him as their pet.

    Huge importance would be read into random bits of dialogue, and I'd have to rework things to make that pay off for them instead of the thing they were supposed to notice.

  • We're subsidising mid-day during-work drinks. This is what Hannah Spencer is talking about - people are showing up to debates and votes three sheets to the wind. Heck, I remember seeing Michael Gove barely able to keep himself upright during the Brexit debates. This is far from the only time he's been seen like this though, and he's by no means the only one.

  • Even funnier, the bars and restaurants are heavily subsidised.

    MPs pay well below the going rate for food and alcohol. The cost to the public is a staggering £5.8 million a year.

  • I'm literally going off the upvotes / downvotes, nothing more.

  • Ah, ok thanks. I've seen this crop up a few times and wondered how the arguments spin wildly out of control. I get it now.

    You (and I assume the OP) are using the word "dangerous" differently than most people here.

    When you say "dangerous", you mean "might be a threat", right? Most people here understand it as "is a threat".

    To use an abstract example, to me "this sandwich is dangerous" and "this sandwich is potentially dangerous" are wildly different. One says this sandwich is definitely poisoned or something, the other is simply telling me to keep my guard up.

  • The post is titled "all men are dangerous".

    Nobody is denying that some men are dangerous. Nobody is denying that you can't tell if someone is dangerous or not. Nobody is denying that men are physically more dangerous to women than other women.

    What you, and everyone else, are saying is that "all men are potentially dangerous".

  • I used to enjoy Penny Arcade when I was younger and much more into the game space. I still have a couple of Penny Arcade t-shirts in the rotation which have got to be over twenty years old by now!

    Back then there was lots going on - starting the Child's Play charity, starting PAX, the whole Jack Thompson saga. Also I got the gaming references in a way that I no longer do...

    I dip into the strip now and then, but don't follow it like I used to. The art has gotten really good now, but I do miss the 2005 style.

  • I'm disappointed but not surprised.

    The government supporting this would be like turkeys voting for Christmas. They'd be exposing the fact that the government is compromised, and the Tories would spin an investigation as evidence of antisemitism.

  • Maybe another thing worth considering is that rectangular flags are just bigger and easier to see than other shapes.

    Also, maybe it just became "normal" to have square flags. The Romans conquered most of Europe, and they flew rectangular banners from their standards. Following the fall of the empire, the different parts of Europe were at war with each other for one and a half thousand years. I suspect all having had this original template, then the subsequent fighting / conquering / reconquering / reconquering, probably lead to this shape becoming normalised.

  • Not just mass-manufacturing - because of how they work, the rectangle is the default shape of cloth produced on any loom.

  • The idea of someone buying up all the grain for something else was the first thing that occurred to me too - essentially the same as with AI and all the GPUs now.

    Maybe for feeding a secret group giants that the crown is using to gain the upper hand in a war.

  • A coworker once told me a trick for knowing where the line is when referring to marginalised groups.

    (Please remember that this is before the big Hamas attack in 2023 led to the current state of affairs).

    His advice, replace the group you're referring to with jews, and reread the sentence. His reasoning is that we have a much better feeling for what's acceptable. So, this would be:

    Is it antisemitic of me to completely filter out jew-related topics from all the media I consume?

    So how to answer?

    If you're blocking a ton of topics that don't interest you, to create a highly personalised feed, then that probably wouldn't seem that weird. If it's the only thing you're blocking, then that might be more easily explained by prejudice.

    At the end of the day, though, you're the only person who knows whether you're transphobic or not. It's easy for someone to say that being indifferent to a marginalised group's suffering is as bad as actually persecuting them, but that ignores the fact that some people are barely keeping it together as it is. We live in horrible dangerous times - your feed is your own business, but I'd suggest trying to keep abreast of the broad strokes of what is happening in the world.

  • I see four white rectangles there. If that's not proof of snow, then I don't know what to tell you.

  • UKCasual @lemmy.world

    MasterChef

  • 3DPrinting @lemmy.world

    FYI, You Can Totally Repair an Extruder with Junk.

  • You Should Know @lemmy.world

    YSK how to talk about the PM of the UK.

  • Fake History Porn @lemmy.world

    India's Chandrayaan-3 Space Mission Successfully Lands at Lunar South Pole (2023)

  • Fake History Porn @lemmy.world

    Photograph Taken During The Fall of the Ming Dynasty (1636)

  • Fake History Porn @lemmy.world

    Ahead of the first human face transplant, a surgeon practices the technique on a gorilla (1977)