Du kannst deine Tweets vom Web aus und über Drittapplikationen mit einem Standort versehen, wie z.B. deiner Stadt oder deinem genauen Standort. Du hast jederzeit die Möglichkeit, Standortangaben nachträglich zu löschen. Mehr erfahren
The closest thing we may have to alien language is neural networks trained to communicate with each other — without any training data.pic.twitter.com/6lSJYB6c92
You can define the kinds of messages you'd like to communicate along with the medium and communication channel, and then train a deep neural network to pass messages through the whole system. This makes it possible to build and experiment with language-generating systems.pic.twitter.com/6JiKFKOVHE
The medium can be anything that a neural network can work with, but I think visual systems are the most interesting. @_joelsimon (left) and @advadnoun (right) have experimented with visual languages.
I imagine that video or even choreography could work, too (bee dance??)pic.twitter.com/pwhHPW5r2O
Here's an example of the first 10 symbols from a visual language-generating system. Each symbol carries a unique message that can be encoded and decoded through a noisy visual channel.pic.twitter.com/FHHej6rAqP
One things that's interested me for a while is the question of if a language-generating system could produce an "visual syntax," something that written languages lack. Natural languages and programming languages both have a tree-like structure, but it has to be parsed.pic.twitter.com/TSgNWkN3BS
This interests me partly because I have no idea what it would look like. Would the patterns be human-interpretable? Would the system learn to draw diagrams, or something totally different? Still actively exploring this, but here's an evocative (but hard to decipher) example.pic.twitter.com/bx5t5Hes9S
Do you have a GitHub repo for this that I could take a look at? Super interesting!
thanks, planning on sharing the code soon!
Thank you for this collection, but I would argue ML does not produce a close-to alien language. Even if the symbols are not human readable, the underlying math could still be comprehended, no? We have lots of alien languages in the animal and plant kingdom.
Here is a silly idea. If the outcome is a visual script, what if you iterate and compare outcome with long lost scripts like harrapan language. If it matches, you would have deciphered a long lost language?
Do you have any code or papers about this you could recommend?
here are two related posts I'd recommend: https://rynmurdock.github.io/2020/02/05/CCN.html … from @advadnoun and https://www.joelsimon.net/dimensions-of-dialogue.html … from @_joelsimon (with code!)
Twitter ist möglicherweise überlastet oder hat einen vorübergehenden Schluckauf. Probiere es erneut oder besuche Twitter Status für weitere Informationen.
