Apple Is Working on a Dedicated Chip to Power AI on Devices • Hacker News Board

Re: Apple Is Working on a Dedicated Chip to Power AI on Devices

#121

Post by dgacmu » Sat, May 27, 2017, 4:07 PM UTC

taneq wrote: post #72

panic wrote: post #69

Earlier quoted context omitted.

Privacy is another reason to keep the computation local.

Good luck getting any of the big players to acknowledge that. >.<

I believe many of them do. Google has TensorFlow Lite: https://techcrunch.com/2017/05/17/googles-tensorflow-lite-br...

Facebook has Cafe2Go. Apple is working on this (and already has bindings optimized to use the ARM vector unit for DNN evaluation).

Running on device, if it can be done with reasonable power, is a win for everyone. Better privacy, better latency, and more robust operation in the face of intermittent connectivity.

Re: Apple Is Working on a Dedicated Chip to Power AI on Devices

#122

Post by GeekyBear » Sat, May 27, 2017, 9:42 PM UTC

bitmapbrother wrote: post #83

npgatech wrote: post #75

Earlier quoted context omitted.

How so? Google developed a server class TPU for Datacenters. Apple is trying to build on-device low powered custom chip.

Google announced it at I/O. They're using DSP's, specialized for neural network processing, on the SoC with TensorFlow Lite on the device. https://youtu.be/Y2VF8tmLFHw?t=1h22m8s

Apple added a neural networking API that can leverage the GPU or CPU of a given device in the last version of iOS.

https://www.wired.com/2016/06/apple-bringing-ai-revolution-i...

There was an article last year discussing customizations Apple had made to Imagination's GPU that made it more suitable for this purpose.

http://www.realworldtech.com/apple-custom-gpu/

Re: Apple Is Working on a Dedicated Chip to Power AI on Devices

#123

Post by yellow_postit » Sun, May 28, 2017, 6:26 AM UTC

visarga wrote: post #95

halflings wrote: post #49

Earlier quoted context omitted.

I don't see how unsupervised learning makes this any better? That data you're training on in an unsupervised manner is still collected somewhere, and could contain as much private information as a labeled dataset.

Yep, whether it's supervised or not has no bearing on privacy. What counts is where the data is processed and who has it.

Labels for supervised training tend to come from humans in the loop. I think many would consider another human looking at their photos, searches, etc. To be a loss of privacy albeit with a small surface area.