You requested someone with a degree in this? *Holds up hand*
nafrondel
1.7K31

I was pondering about this “dog=0x3b chien =0x3b perro=0x3b. Where 0x3b is the idea of dog.”-thing and I can relate to the logic as a human. But what the AI is missing, which all the humans possess, while forming those “ideas”, is visual perception. How can the AI learn the meaning of a word without having a secondary input to compare against? Our idea of a dog is a visual one, an animal with 4 legs and inherent characteristics. So when we see one in China, we know it is a dog without knowing the word in Mandarin. How does an AI formulate “an idea” when it only has one kind of input? This escapes me.
It looks like the Google AI still made up this “ox3b” as a bookmark to the cloud of “dogs” in all the languages it has learned, hence accessing it faster and more reliably. Or do we think if we describe the animal in enough detail to it, it can correctly suggest that we are talking about “the dog”?

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.