I definitely recommend switching to Python. Everything Keras related is being developed for Python first.
@fchollet
-
Universal API for ML Components Across All Frameworks
By
–
You can use this API to write layers, metrics, optimizers, etc. that work the same in any framework. One codebase, compatible with the entire ML ecosystem.
-
NumPy optimization framework supporting multiple deep learning backends
By
–
You can think of it as an optimized & extended NumPy that compiles to any of JAX, TF, PyTorch, or… NumPy itself.
-
Keras 3 Introduces Low-Level API Supporting JAX TensorFlow PyTorch
By
–
Keras 3 features a low-level `keras.ops` API with 201 functions that work in JAX, TensorFlow, and PyTorch with the exact same semantics and numerics. This includes 137 NumPy ops. It's not "numpy-like", it's literally the NumPy API (plus a bunch of NN ops).
-
AI Acceleration Drives Progress in Scientific Research
By
–
Accelerating the development of AI will help accelerate scientific progress across the board.
-
Intelligence as Behavior Under Uncertainty and Change
By
–
Intelligence is about generating adequate behavior in the presence of high uncertainty and constant change. If you could have full information and if your environment were static, then there would be no need for intelligence — instead, *compression* would give you an optimal
-
Intelligence and Compression: Opposite Nature in Generalization
By
–
While intelligence *leverages* compression in important ways in representation learning, intelligence and compression are by nature opposite in key aspects. Because intelligence is all about *generalization to future data (out of distribution)* while compression is all about
-
Current AI Systems Achieve Superhuman Skill in Narrow Well-Defined Tasks
By
–
It's well established that current systems, trained only on human generated data, can achieve superhuman skill (super-humanity skill as per OP) as long as the target task is sufficiently narrow and well defined. The problem is generality not skill.
-
GPT-4 AGI Claims Fade as Monthly AI Hype Cycle Continues
By
–
Last Fall, the word in SF was that GPT-4 was already AGI. Etc etc. Now it's just a monthly thing.
-
Historical AGI Panic Cycles: From Q-Learning to Deep Reinforcement Learning
By
–
The first panic over imminent AGI was circa 2013 about Atari Q-learning by DeepMind. The second one was circa 2016 over Deep RL (partially triggered by AlphaGo). So many folks in late 2016 were convinced that Deep RL would lead to AGI in under in 5 years…