Well holy shit, this is cool.

[toxic_flange@infosec.exchange]

Well holy shit, this is cool.

Someone reverse engineers Apples' Neural Engine and trained a neural network on it.

Sure I shit on AI usage today, but its general corporate backed and pushed usage and the general snake-oil salespeople enablers.

This is a great combination use of LLM to reverse engineer Apple's practically undocumented neural engine and make it do shit Apple didn't want you to do.

I see Apple's AI related stock going up vs Nvidia

• M4 ANE = 6.6 TFLOPS/W vs 0.08 for an A100 (80× more efficient)
︀︀• "38 TOPS" is a lie - real throughput is 19 TFLOPS FP16
︀︀• Your Mac mini has this chip sitting mostly idle
︀︀
︀︀Translation: local AI inference that's faster AND uses almost no power.

Inside the M4 Apple Neural Engine, Part 1: Reverse Engineering

How we bypassed CoreML and talked directly to the hardware

(maderix.substack.com)

GitHub - maderix/ANE: Training neural networks on Apple Neural Engine via reverse-engineered private APIs

Training neural networks on Apple Neural Engine via reverse-engineered private APIs - maderix/ANE

GitHub (github.com)

#ai #LLM #apple #hardware