Cabel Sasserthis graduation speech moment is notable, and her amazed shock at having failed to read the room feels instructive.
when you’re inside the bubble, you think everybody else is. but everybody isn’t.
myrmepropagandistI can't imagine "the internet" getting boo'd like that in 2001 Grads would have cheered along for "internet"
Or even like bitcoin in say 2010, lots of people were skeptical but would not have just boo'd
This is remarkably unpopular.
Landa 
@futurebird
Proponents are always comparing it to the industrial revolution, but maybe it's better compared to the likes of leaded gasoline, CFC aerosol cans, or asbestos anything.
Maybe some of us have learned to spot a pattern.
@cabel
anonymousplI was going to say "no asbestos is useful" but thinking about that more I think you are correct, because so are LLMs in a very narrow setting, just like asbestos, but instead we have foolish business persons who want to put this stuff in everything.
bencourtice@futurebird @anonymouspl @[email protected] @cabel when I did asbestos handling training, they explained that it didn't get banned for lack of uses - it's very useful! - but because it happens to kill people horribly. A geologist friend was recently diagnosed with mesothelioma after workplace exposure many years prior. It's still killing people.
PalmAndNeedle@bencourtice @futurebird @anonymouspl @cabel asbestos is awesome, from a material science perspective
@PalmAndNeedle @futurebird @anonymouspl @cabel yeah. As long as someone else has to actually work with it!
Ratsnake@bencourtice @PalmAndNeedle @futurebird @anonymouspl @cabel so what you are saying is we should make AI handle asbestos
@ratsnakegames @PalmAndNeedle @futurebird @anonymouspl @cabel Yes! we can bury both of them in the same registered landfill together
We Must DissentAnother material that is just taunting us is lead. Basically perfect option for so many use cases but so, so poisonous
Noodlemaz@bencourtice @futurebird @anonymouspl @cabel yep, since it takes decades for the cancer to develop, we're now seeing the impact on people who were installing it in the 60s-80s.
It's horrific, very hard to treat, most won't survive. We didn't know at the time, is the thing.
But we already have evidence that LLMs can kill people too. The absolute lack of giving a shit is overwhelming.
robertogonzalez@noodlemaz @bencourtice @futurebird @anonymouspl @cabel I'm afraid we did know about asbestos way before the 60's. First lawsuit is dated in 1929. We just didn't care enough. Same thing happened whenever a profit is in sight, see GHG as a quite prominent example.
Feisty@robertogonzalez Maybe we will see Make Asbestos Great Again. I don’t even think that is far-fetched at this point. @noodlemaz @bencourtice @futurebird @anonymouspl @cabel
cuan_knaggs@futurebird purpose built llms with a specific defined focus can be game changing. llm-as-a-service is useful for spam generation @anonymouspl @Landa @cabel
opotamus@futurebird @anonymouspl @Landa @cabel
True. As well as killing off their own consumers and workforce, but at a slower rate than those foolish business persons' short term profit windfall so they ignore it, also like asbestos.
Plus those foolish business persons will never properly be held accountable, also like asbestos.
ROTOPE~1 ⭐️
Elizabeth 
@rotopenguin @futurebird Did someone show asbestos? I'm not sure if this is better or worse, but, I always enjoy this stuff.
Jennifer Moore 😷Oof!
What's the source of the photo, please?
(would like to save it if it's creative commons, and/or credit the photographer - also curious about the vintage)
Mensch, Marina
StarkRG@futurebird @anonymouspl @Landa @cabel LLMs aren't useful at all, they only *seem* useful. Paper straws are more useful than LLMs because, although they don't last very long and aren't recyclable at all (unlike plastic straws, which are slightly recyclable), they *do* adequately perform the task they're put to.
LLMs are like very realistic wax fruit, they do an incredible job at looking like they're just as good as the real thing, but they definitely are not.
Ian Tindale@StarkRG @futurebird @anonymouspl @Landa @cabel LLMs are highly useful in the field of trying to translate animal communication into something we can understand, and significant advances have been made in this area – this in turn will allow us to stand a chance of having the slight possibility of maybe being able to have a system in place to attempt to communicate with aliens, when they land (once we’ve hidden all our QWERTY keyboards or they’ll piss themselves laughing). That’s about the only honest use for them that I can see, though.
@u0421793 @futurebird @Landa @cabel @anonymouspl
Don't take this as me saying you're wrong, but I don't really see how something which contains no information about meaning of our own languages could be used to uncover the meaning of an unknown language. LLM based translations are already very not good, and that's between languages we already have translations for and are all created by human minds besides.
Don't take this as me saying you're wrong, but I don't really see how something which contains no information about meaning of our own languages could be used to uncover the meaning of an unknown language. LLM based translations are already very not good, and that's between languages we already have translations for and are all created by human minds besides.
muniatu endormiscat@anonymouspl @u0421793 @cabel @Landa @futurebird @StarkRG
LLMs can be really good at recognizing patterns. They don't really uncover the meanings of an unknown language, they just make the patterns much easier for us to recognize (for animal sounds, things like "this is where a word ends" and "this is a different sentence"). And we are the ones to try to uncover the meanings, since we are the ones who have a knack for ascribing meaning to patterns
LLM based translations are bad because the LLMs are forced to try to ascribe the meanings themselves, from "patterns" that are already plainly written (as words), and then rewrite the meanings into a different set of patterns. It's the opposite work
This is similar to how they can be a great tool for medicine research too
@soc
I can see the underlying architecture, deep learning transformers, being useful in quite a lot of fields, including medical research, and that's what it sounds like you're taking about. LLMs, though, large language models, don't seem to have much real use. I even doubt they'd be of much use to linguists studying a language since all of its data gets stored in incomprehensible and mostly inaccessible ways.
@anonymouspl @u0421793 @cabel @Landa @futurebird @StarkRG
Sorry, yeah, my bad. I was indeed talking about LMs, without the "Language", just "Large Models". The extra L accidentally slipped in bc of habit
Edit: Most people aren't aware or know to distinguish between LLMs and LMs either, so that's why I assumed that's what Ian was talking about too
@[email protected] @futurebird @StarkRG @Landa @cabel @anonymouspl Also, don’t fall into the trap of thinking that non-human language is just a bunch of other different sounds made by animals. I anticipate much more advanced ways of capturing, recognising and digitising language of not-humans, which might end up placing more emphasis on not-sounds, eg stance, position relative to things, timing, etc. This would still go into a tokenisation funnel the same way that a novel verbal language would be, for the trainings.
@[email protected] @futurebird @StarkRG @Landa @cabel @anonymouspl ultimately it’s all about the embeddings - does a particular ‘thing that happened’ have any meaning to other things that recognise that particular thing that happened. In a way it doesn’t have to be animals or plants or living things. Perhaps geological and weather and stellar events can be tokenised as a set of embeddings with weights and stuff, and this might reveal patterns which ‘mean something’ which otherwise are occluded because we just don’t ‘see’ the relationships. Is this language? maybe, but it’s not from life. Or maybe we should consider the sequences of variances of situations on our planet to be the communications of a planet that is ‘alive’ – it’s life, but not as we know it.
@u0421793 I haven’t heard of that. A link to an overview would be much appreciated. Thank you! :)
@Landa there are a few initiatives currently:
• https://cacm.acm.org/news/can-ai-talk-to-the-animals/
• https://www.nature.com/immersive/d41586-024-04050-5/index.html
• https://earthspecies.org/
• https://blog.google/innovation-and-ai/products/dolphingemma/
• there’s others on Bonobos, Elephants and Rodents…
• https://en.wikipedia.org/wiki/Animal_communication (Not specifically mentioning LLMs but it does cover the range of communication modes in animals that convey messages, and these would have to be observed, acquired and digitised and then tokenised as Q K and V matrix/weights from training)
• https://cacm.acm.org/news/can-ai-talk-to-the-animals/
• https://www.nature.com/immersive/d41586-024-04050-5/index.html
• https://earthspecies.org/
• https://blog.google/innovation-and-ai/products/dolphingemma/
• there’s others on Bonobos, Elephants and Rodents…
• https://en.wikipedia.org/wiki/Animal_communication (Not specifically mentioning LLMs but it does cover the range of communication modes in animals that convey messages, and these would have to be observed, acquired and digitised and then tokenised as Q K and V matrix/weights from training)
The Sleight Doctor 🃏🍉@anonymouspl @Landa @futurebird @cabel I've started comparing it to universal acid, because it eats away at and degrades the quality of everything it touches. 🤷🏻♂️