You no longer need to leave Python to write high-performance hardware kernels. Learn how to use Pallas in Keras to author custom ops that lower to Mosaic for TPUs or Triton for GPUs:
@fchollet
-
Keras API: Accessing JAX’s Full Power Without Reinventing
By
–
The point of the Keras API is to let you tap into everything JAX has to offer but without reinventing the wheel
-
AI Offloading Enables Cognitive Performance Across All Ages
By
–
One positive outcome of AI is that it will make the aging-related decline of your brain power less relevant. You can keep doing the same things just as nimbly at any age by offloading more.
-
Code as Liability: Why Generating More Code Isn’t Always Better
By
–
I tend to view code as more of a liability than an asset. In this light, making it cheaper and faster to generate a lot of code might not be an unmitigated blessing.
-
GenAI Raises the Bar for Mediocrity, Making Competence Economically Worthless
By
–
GenAI will not replace human ingenuity. It will simply raise the floor for mediocrity so high that being "pretty good" becomes economically worthless.
-
LLMs Give the Impression of Knowing More Than You Really Do
By
–
LLMs are great at giving you the impression that you know more than you really do, and of course they will always praise your crazy ideas
-
LLMs and the Illusion of Scientific Breakthroughs
By
–
When it comes to scientific discovery, one thing LLMs are really good at is getting hobbyists to delude themselves into believing they've made a huge breakthrough on some longstanding problem or a theory of everything
-
LLMs as Brainstorming Partners for Scientific Discovery and Innovation
By
–
Evaluating the potential of LLMs to help with scientific discovery. In short: new ideas are direly needed to move AI towards invention. LLMs can be useful as brainstorming partners though.
-
Keras 3.13 Release: LiteRT Export, GPTQ Quantization, Adaptive Pooling
By
–
New Keras release: 3.13 Some major new features: • Model export to LiteRT (formerly TFLite) for mobile/edge
• GPTQ quantization support for post-training compression
• New Adaptive Pooling layers for dynamic architectures -
Differentiable Programming: Representation vs Learnability Challenge
By
–
The ability to embed your discrete function on a differentiable curve is a prerequisite for differentiable programming, but it is *far* from sufficient, the hard part is getting GD to find the right function Anything can be represented, not everything can be learned