Deep Learning Is Hitting a Wall

"What would it take for artificial intelligence to make real progress?" Classic case of missing the forest for the trees.

I can appreciate the author is trying to highlight current challenges in an attempt to start a discussion, but having witnessed the very real progress ML has made in just the last 10 years, it's hard not to read that subtitle as a bit inflammatory. With every new breakthrough - or even moderate improvement - the goalposts of what constitutes ML/AI progress are shifted just far enough to buy the naysayers just enough plausible deniability that they aren't immediately dismissed when they point out that "sure, ML has given us Siri/Alexa, StableDiffusion, AlphaZero, image recognition and self-driving cars that don't immediately crash, but what about cancer diagnosis and AGI?"

I wonder what the world of the Cambrian Explosion would look like to those naysayers if they had been around at the time. "Look at all these evolutionary dead-ends. Locomotion? Nervous systems? Skeletons? All this trial and error will never amount to anything. We need to make real progress."

Deep learning (research) has hit a wall. Hint hint no more funding.

Deep learning applications are just getting started and from what I'm seeing on the ground, will take no less than 5-10 years before we see it take off.

Archived in full: https://archive.is/WvGoZ

Hitting a wall??? Deep learning is winning and making insane strides right now. What is this nonsense?

The only wall I see is a paywall.

can't understand why something is learning, or what it's learned, then you can't fix it when it stops learning.