The use of software in chip design is not new.
And it does it in a fraction of the time it would take a human to do so.
The AIs superiority to human performance has drawn a lot of attention.
It’s free, every week, in your inbox.
Basically, what you want to do is place the components in the most optimal way.
This is the manifestation of one of the most important and complex aspects of human intelligence:analogy.
While we take these skills for granted, theyre what makes us very good at transfer learning.
But the scientists faced a problem that was orders of magnitude more complex than Go.
(greater than102,500), whereas Go has a state space of10360, the researchers wrote.
The chips they wanted to design would be composed of millions of nodes.
The term intuition is often used loosely.
But it is a verycomplex and little-understood processthat involves experience, unconscious knowledge, pattern recognition, and more.
Fortunately, putting these intuitions to test is becoming easier with the help of high-power computing andmachine learning tools.
Its also worth noting that reinforcement learning systems need a well-designed reward.
The weights are hyperparameters they had to adjust during the development and training of the reinforcement learning model.
Curated datasets
The deep neural online grid used in the system was developed usingsupervised learning.
Supervised machine learning requires labeled data to adjust the parameters of the model during training.
To avoid manually creating every floorplan, the researchers used a mix of human-designed plans and computer-generated data.
But without quality training data, supervised learning models will end up making poor inferences.
We humans use all kinds of shortcuts to overcome the limits of our brains.
We cant tackle complex problems in one big chunk.
But we can design modular, hierarchical systems to divide and conquer complexity.
Ill give an example of software engineering, my own area of expertise.
But software developers never write their programs that way.
We then nest those pieces into larger pieces and gradually create a hierarchy of components.
You dont need to read every line of a program to understand what it does.
Modularity enables multiple programmers to work on a single program and several programs to reuse previously built components.
We often trade speed for modularity and better design.
After a fashion, the same can be seen in the design of computer chips.
Human-designed chips tend to have neat boundaries between different modules.
Those are the kinds of skills that better AI chips can enhance but not replace.
If theres a virtuous cycle, its one of AI and humans finding better ways to cooperate.
you could read the original articlehere.
