"Humans are conscious; this big curve fitted on tons of human-generated outputs can reproduce human-like behavior in some cases; therefore this big curve is conscious" has got to be some of the most mindless, most hubristic reasoning I've ever seen.
@fchollet
-
Expert Claims Must Be Supported by Identifiable Evidence
By
–
Also, reminder that if an expert asserts "after looking at this thing, I can conclude X", whether that expert is a human or a machine, they should be able to point out what evidence they found and why it is good evidence. If they can't do that, don't trust them.
-
ML Systems Cannot Magically Recover Hidden Information
By
–
A prevalent cognitive fallacy around ML systems is the belief that they can magically recover hidden information that is out of reach for human experts — "the AI detected that those Apollo pictures are fake!", "the AI can predict if the person in the picture is a criminal!",
-
Community Rapidly Adopts PyTorch and JAX Backends
By
–
Really impressive how quickly the community has been adopting the PyTorch and JAX backends!
-
Technology growth and the tension between VC interests and social equity
By
–
But the thing is, people want technology to contribute to a kind of growth where benefits would be broadly shared. Socially aware growth. This is often at odds with what VCs advocate for.
-
Tech Invents Imaginary Threats Instead of Real Problems
By
–
The tech community clearly doesn't have enough real threats, because it seems to keep coming up with imaginary ones — from "AI x-risk" to "decelerationists", a group that supposedly opposes progress and advocates for planned decline.
-
Image Classification with Global Context Vision Transformer Tutorial
By
–
New tutorial on http://
keras.io: Image classification using Global Context Vision Transformer. Works with JAX/TF/PyTorch. -
JAX Framework: Transform Your ML Development Expectations
By
–
Anyway, if you haven't explored JAX yet, you really should. It makes you completely change your expectations for ML frameworks 🙂
-
Out of the Box Performance Without Optimization Efforts
By
–
It's representative of "out of the box performance", the speed folks who aren't performance experts get when they develop their own models. We have made zero performance optimizations to our model. Also we have found torch compile() rarely ever works, and when it does it may not