Discovering the world of artificial intelligence with Axiom Zen.
Androids Dream of Uncanny Sheep
Moore's Law has been proven wrong, but artificial intelligence continues making huge strides in chip technology. Plus, sheep's pain costs big bucks, autism is taken down a peg, and drug manufacturers hope to cure 100 diseases in only 10 years.
Article of the Week
How AI Can Keep Accelerating After Moore’s Law
"Researchers had figured out how to automate some of the work of crafting machine-learning software, something that could make it much easier to deploy the technology in new situations and industries... [But] the project had tied up hundreds of the chips for two weeks solid—making the technique too resource-intensive to be more than a research project even at Google."

Moore’s Law forecasted that the number of transistors that could be fitted into a given area of a chip would double every two years. That acceleration has now slowed, but instead chip manufactures are making chips that are more specialized for AI tasks, which is helping them overcome the hurdle they would otherwise be facing.
The Bleeding Edge
"The program’s average accuracy at identifying [facial action units] (which are associated with different levels of pain) was 67%, about as accurate as the average human."

Why do we care if sheep are in pain? Well, farmers care because knowing if a sheep is in pain helps them treat injury and infection that might otherwise be lethal. Artificial intelligence researchers care because this image recognition has a wide range of applications, from farming to education and even customer service.
A Breakthrough AI Can Now Predict Which Babies Will Develop Autism
"The method has been applauded for its non-intrusive nature, ability to identify autism from only one scan, and potential to increase the feasibility of developing early preventative interventions for ASD. ”This is a game-changer for the field.”"

Historically, it's been impossible to engage with preventative intervention for autism, because we haven't been able to predict which children will be affected by it. But now, we can use machine learning to predict which siblings of autistic children will develop the disorder, and hopefully we will soon be able to apply that knowledge to other children.
"Is it possible to find cures for 100 rare diseases in a decade or less? Yes, says Blake Borgeson, CTO of Recursion Pharmaceuticals."

We won't know for another ten years if they're going to be successful, but the idea is a noble one. The company is using machine learning to amalgamate research in order to apply the same cures to different diseases, and make each scientist's individual work go further. Using an open source visualization framework that operates on a Python platform, scientists can have the computers help with data visualization and other key components in drug manufacturing.
Now Artificial Intelligence Has Invented Sounds That Have Never Been Heard Before
"AI is generating brand new sounds that have never been heard before, and they could soon be coming right out of your radio."

Can you imagine a colour you've never seen before? How about a sound that's never been heard? Unfortunately that is nearly impossible, but what we can do, thanks to a clever application of machine learning, is blend existings sounds together to create something mostly new. Musicians will now be able to blend instruments and create melodies that would have been impossible with an real world instrument, all from the comfort of their computer screens.
"The pix2pix project demonstrates something pretty profound about machine learning circa 2017: It's awful at generating new images, or at least meaningful new images."

Remember a few months ago, we found this cute program where you could draw a cat, or a purse, or a house, and it would create a "photo" based on that sketch? Well, they've moved on to human faces, and the results are about as terrifying as you would expect.
“By far the greatest danger of artificial intelligence is that people conclude too early that they understand it.” ― Eliezer Yudkowsky
AI@AZ · #350-980 Howe Street · Vancouver BC V6Z 0C8 · Canada
Unsubscribe