Machines, Lost in Translation: The Dream of Universal Understanding

It seems to me that we're still waiting on a breakthrough. FTA:

> "There's no saying where machine translation can go when the electronic brain meets the human one."

The neural networks that we modeled after our brains have recently become much faster, and subject to much more input (training data) thanks to our information age, but has the network fundamentally changed?

I foresee something of a plateau with the current model. Yes, in recent decades we've seen exponential increases in MT accuracy and speed, but it's still missing so much in terms of contextual relevance, and tiny errors/hesitations in human thinking/speaking.

Yes, we can improve the algorithms and input processing speed, but are we really on the brink of a breakthrough of something better? I think that remains to be seen.