@cat_abyss a related bit of Canadian content, that always makes me smile, is this short poem entitled
by Leonard Cohen
from The Book Of Longing:
Out of the thousands
who are known,
or who want to be known
maybe one or two
and the rest are fakes,
hanging around the sacred precincts
trying to look like the real thing.
Needless to say
I am one of the fakes,
and this is my story.
@Nafrondel But, yeah, the fear of them becoming evil also comes up a lot. I have a looot to say about that, but the short version is that an artificial consciousness of advanced capabilities is unlikely to care we exist. Humans just aren't that important or interesting or useful really. My friend Damien writes about ethics and machine learning a fair bit if you're into that. You can find some of his work here:
I think he's decidedly excellent.
@Nafrondel oh, yeah, but I've actually had people (including a machine learning prof criticizing neural nets) tell me that story as being about using ANNs for classification trained on photos of enemy tanks but the ANN learns to recognize sunny days because all enemy tank photos were sunny and all the other photos were on cloudy days. I don't know how it mutated from what you've said into this other story, but it has been floating around in that form for decades.
@Nafrondel it's so weird to me. Why is this story of ANNs as black boxes so persistent? People even sometimes still tell that old tale about the military tanks that were accidentally trained to only fire on sunny days. Despite there being no evidence that any such thing ever happened and it now being recognized as apocryphal. It's not magical gnomes inside; it's numbers. We can inspect numbers.
In the end it's just matrices of numbers (representing weighted connections and biases) and known activation functions that we selected before training (usually Gaussian or sigmoidal but anything differentiable will do if you like to experiment). Simple? Probably not. Utterly impenetrable? No.
Maybe that made no sense without context.
So, here’s the thing about understanding neural network solutions:
I’ve done it.
I’ve trained networks and then interpreted their internal structure to figure out how they solved the problem of getting from point A to point B.
A big big caveat is that I’ve done this with fairly small networks.
But I’m keen to figure out how to scale things up.
My concern is that people keep hearing it is impossible and as a result very few of us try.
Welp, I've contacted a favourite podcast to correct their statement that neural networks are impossible to understand. I'll do more of this in the future, I think, but I'll probably wait until I'm done the paper I'm currently writing since it has some new approaches not in my previous publications.
Shamelessly favouriting every cat photo I see. ❤️🐱❤️
FYI - if you assume foul intent because I'm white/straight/CIS, I totally get it.
Like, you have every right to be wary of our shit because we fuck stuff up all the time. It's almost all we do.
"I have become a language warrior. I want to join all those others in the world who are fighting for marginalized languages. No language is ever marginal to the community that created it. Languages are like musical instruments. You don’t say, let there be a few global instruments, or let there be only one type of voice all singers can sing."
-Ngũgĩ wa Thiong'o
Rubbing fuzzy tummies.
Life is pretty grand.
also: the German term for the wind in your face you get from moving at speed is "Fahrtwind"
do with this information what you will
How to get a private 🔒 toot to your followers on a different instance without mentioning them: if you have an account on that instance, mention it! That way, it federates and is readable by your followers there!
Wie du privat🔒einen Toot an Followis auf einer anderen Instanz machst, ohne sie zu mentionen: wenn du einen Account auf der anderen Instanz hast, mentione den! So föderiert der Beitrag und ist von Followis auf der anderen Instanz lesbar.