Machines making music, translating Chinese, self-driving trucks, and more Welcome to this week's AI roundup. We have news on a machine learning model used by Google to make music that doesn't sound completely bad, improved translation between English and Chinese from Microsoft, and a new test bed for Waymo's self-driving trucks. Making music with machine learning - Google researchers have developed …
Typical Song Structure: Intro, verse, chorus, verse, chorus, bridge, verse, chorus, ending.
The little melody bits are used to populate these. It's almost trivial. Assembling a passable song only takes a few hours of fiddling on the computer; yes I've done it. Surprisingly easy.
Making a Hit requires multiple skills, vast luck and expensive promotion. I have none of these.
I think I learned more about sound and sine waves from playing around with an old analog synthesizer as a kid than I could ever learn from a school or a book.
My friend and I would compete with each other by thinking up everyday sounds and seeing how fast the other could produce it on the synth by tweaking the oscillators, filters, pitch wheels and varying speed of sound pulses.
BTW, that short sound clip was pretty good in IMHO.
Lol, :-)
This list of funny Chinese to English translations pictures has grown bigger into a huge collection of 140 photos grouped in 10 typical scenarios one could encounter when travelling in China. The restaurant, the hotel, and toilet, the park... Check it out at https://www.actranslation.com/knowledge/fun/hilarious-translations.htm.
There is still a long way to go for the AI.
"...you can also play around with algorithms that create new neural networks."
Instead of wasting your lives manually experimenting with various algorithms and settings used to optimally tweak the Machine Learning / neural net, wrap that very optimization process itself in one more layer of automated Machine Learning.
Do this first with many medium size systems to quickly explore the concept while gathering results, then double-check with several huge examples to spot check that it still applies to larger systems.
Let it run experiments for a month, and then run to the Patent Office with the likely-optimal algorithm and a table of its optimal settings for various classes of problems.
Small team + a few months = the definitive paper on the subject.
Surrounded by advanced computers, manually running experiments. Yeah, A.I. is hard.
Google's GNMT is indeed making great progress in its English>Chinese translation performance, but sometimes it still makes funny mistakes that would only leave people in laughter. Many westerns who had travelled in China would know what I am talking about - they must have been amuzed / confused by some of the machine translated road signs or restaurant menus - such as the famous "husband and wife lung slices" (actually meaning "beef slices")...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
