Funnily enough it isn't even required by their purported bayesian doctrine (which proves none of them do the math), you could simply "update forward" again based on the new evidence that the text is part-fictional.
zogwarg
joined 2 years ago
Counter-theory: The now completely irrelevant search results and the idiotic summaries, are a one-two punch combo, that plunges the user in despair, and makes them close the browser out of disgust.
Subjectively speaking:
- Pre-LLM summaries were for the most part actually short.
- They were more directly lifted from human written sources, I vaguely remember lawsuits or the threat of lawsuits by newspapers over google infoboxes and copyright infringement in pre-2019 days, but i couldn't find anything very conclusive with a quick search.
- They didn't have the sycophantic—hey look at me I'm a genius—overly-(and wrong)-detailed tone that the current batch has.
This is obviously a math olympiad gold medal performance, Fields medal worthy even!
It can't be that stupid, you haven't read the sequences hard enough.
I mean if you want to be exceedingly generous (I sadly have my moments), this is actually remarkably close to the "intentional acts" and "shit happens" distinction, in a perverse Rationalist way. ^^
But code that doesn’t crash isn’t necessarily code that works. And even for code made by humans, we sometimes do find out the hard way, and it can sometimes impact an arbitrarily large number of people.
Did you read any of what I wrote? I didn't say that human interactions can't be transactional, I quite clearly—at least I think—said that LLMs are not even transactional.
EDIT:
To clarify I and maybe put it in terms which are closer to your interpretation.
With humans: Indeed you should not have unrealistic expectations of workers in the service industry, but you should still treat them with human decency and respect. They are not their to fit your needs, they have their own self which matters. They are more than meets the eye.
With AI: While you should also not have unrealistic expectations of chatbots (which i would recommend avoiding using altogether really), it's where humans are more than meets the eye, chatbots are less. Inasmuch as you still choose to use them, by all means remain polite—for your own sake, rather than for the bot—There's nothing below the surface,
I don't personally believe that taking an overly transactional view of human interactions to be desirable or healthy, I think it's more useful to frame it as respecting other people's boundaries and recognizing when you might be a nuisance. (Or when to be a nuisance when there is enough at stake). Indeed, i think—not that this appears to the case for you—that being overly transactional could lead you to believe that affection can be bought, or that you can be owed affection.
And I especially don't think it healthy to essentially be saying: "have the same expectations of chatbots and service workers".
TLDR:
You should avoid catching feelings for service workers because they have their own world and wants, and it is being a nuisance to bring unsolicited advances, it's not just about protecting yourself, it's also about protecting them.
You should never catch feelings for a chatbot, because they don't have their own world or wants, it is cutting yourself from humanity to project feelings onto it, it is mostly about protecting yourself, although I would also argue society (by staying healthy).
Don't besmirch the oldest profession by making it akin to souless vacuum. It's not even a transaction! The AI gains nothing and gives nothing. It's alienation in it's purest form—no wonder the rent-seekers love it—It's the ugliest and least faithful mirror.
✨The Vibe✨ is indeed getting increasingly depressing at work.
It's also killing my parents' freelance translation business, there is still money in live interpreting, and prestige stuff or highly technical accuracy very obviously matters stuff, but a lot of stuff is drying up.
Jinsatsu Zetsubō (人殺・絶望, but his thralls call him Ginny) was not your ordinary vampire goth demon lord... He delighted in his garments of true terror and dread, what better source of inescapable despair than his beige ulster coat, barely held together by off-yellow gold pins, with a salmon pink napkin in the over pocket, an ensemble designed to inspire trudgery sucking all soul and joy from any passerby...
59
AGI Sparklings proponents rejoice! Finding a literal map(*) means LLMs have a world model.
(awful.systems)
Transcribed:
Max Tegmark (@tegmark):
No, LLM's aren't mere stochastic parrots: Llama-2 contains a detailed model of the world, quite literally! We even discover a "longitude neuron"Wes Gurnee (@wesg52):
Do language models have an internal world model? A sense of time? At multiple spatiotemporal scales?
In a new paper with @tegmark we provide evidence that they do by finding a literal map of the world inside the activations of Llama-2! [image with colorful dots on a map]
With this dastardly deliberate simplification of what it means to have a world model, we've been struck a mortal blow in our skepticism towards LLMs; we have no choice but to convert surely!
(*) Asterisk:
Not an actual literal map, what they really mean to say is that they've trained "linear probes" (it's own mini-model) on the activation layers, for a bunch of inputs, and minimizing loss for latitude and longitude (and/or time, blah blah).
And yes from the activations you can get a fuzzy distribution of lat,long on a map, and yes they've been able to isolated individual "neurons" that seem to correlate in activation with latitude and longitude. (frankly not being able to find one would have been surprising to me, this doesn't mean LLM's aren't just big statistical machines, in this case being trained with data containing literal lat,long tuples for cities in particular)
It's a neat visualization and result but it is sort of comically missing the point
Bonus sneers from @emilymbender:
- You know what's most striking about this graphic? It's not that mentions of people/cities/etc from different continents cluster together in terms of word co-occurrences. It's just how sparse the data from the Global South are. -- Also, no, that's not what "world model" means if you're talking about the relevance of world models to language understanding. (source)
- "We can overlay it on a map" != "world model" (source)
@ESYudkowsky: Remember when you were a kid and thought you might have psychic powers, so you dealt yourself face-down playing cards and tried to guess whether they were red or black, and recorded your accuracy rate over several batches of tries?
|
And then remember how you had absolutely no idea to do stats at that age, so you stayed confused for a while longer?
Apologies for the usage of the japanese; but it is a very apt description: https://en.wikipedia.org/wiki/Chūnibyō,
And of course no experiments whatsoever, the cost of the Manhattan project, the hundreds of thousands of employees were merely a "focusing" magick, a sacrifice to re-enforce the greater powers of our handful of esteemed and glorious thinking men, who wrought the power of destruction from the æther.
@ESYudkowsky: Yes, but because the first nuclear weapon makers knew what the duck they were doing - analytic precise prediction of desired outcomes and of each intervening step. AGI makers lack similar mastery or anything remotely close, and have a much harder problem; that's the big issue.
@EigenGender: seems pretty noteworthy that the first nuclear weapons were made under conditions where they couldn’t do any experiments and they involved a lot of math but still worked on the first try.
view more: next ›
They're just very dedicated to the bit... right?