BRN (ASX) ARM teams up with Intel to make use of its 22nm nodes for better smartphone chipsets

Originally posted by gwood888: ↑

Let's look at the supposed leading Edge AI chip:

The Qualcomm® Neural Processing SDK for artificial intelligence (AI) is engineered to help developers save time and effort in optimizing performance of trained neural networks on devices with Snapdragon. And as part of our Qualcomm AI Stack, it can help add a neural network on device without a need for connection to the cloud.Qualcomm Neural Processing SDK for AIPremium tier Snapdragon® mobile platforms have extensive heterogeneous computing capabilities that are engineered to allow the running of trained neural networks on device without a need for connection to the cloud. The Qualcomm Neural Processing SDK is designed to help developers run one or more neural network models trained in Caffe/Caffe2, ONNX, or TensorFlow on Snapdragon mobile platforms, whether that is the CPU, GPU or DSP.
The Qualcomm Neural Processing SDK provides tools for model conversion and execution as well as APIs for targeting the core with the power and performance profile to match the desired user experience. The Qualcomm Neural Processing SDK supports convolutional neural networks and custom layers.The Qualcomm Neural Processing SDK does a lot of the heavy lifting needed to run neural networks on Snapdragon mobile platforms, which can help provide developers with more time and resources to focus on building new and innovative user experiences.As you can see, the Qualcomm's Snapdragon does not appear to have a "true" Edge AI platform and only "supports convolutional neural networks".

So for those who think Mercedes has abandoned Brainchip's Akida and has gone with Qualcomm's Snapdragon for Edge AI, don't despair...Mercedes will still need Brainchip's Edge AI platform
I'm not sure how previous versions of Snapdragon in the last several years have still claimed Edge AI functionality without an actual "true" Edge AI platform...is it all smart software processing?...interesting!
My simple take is that Akida will work with the sensors for learning and inference which creates some kind of "model file" to pass through Snapdragon's chip (or Intel's etc..), which is further optimizes the inference to produce the AI function.
Will Akida become a layer in the Qualcomm AI stack?
Is it easier to incorporate Akida into the stack as an IP or another Edge AI platform as another chip?
I have no technical answer to this question but there is an obvious logical answer.

Brainchip just need to stay its course of R&D to continue to have the most advanced AI neural network that's adaptable with any chip that a Qualcomm, Intel, ARM or Samsung etc..will produce and success will certainly be on the horizon...Akida AKD1500 is certainly another leap ahead of competitors!

Expand

Qualcomm is basically following Apple and Google's lead when it comes to having a dedicated AI processor on their phone SOCs.

But it's not edge neuromorphic AI - it's a processor that can take ML algorithms from a server and process them on board the device. That doesn't make it a terrible bit of gear, my experience with the Pixel 6 Pro is fantastic, but it's not the same as Akida.

The important thing to understand is that Akida could eventually take the place of the dedicated ML processor on these devices. The ML processors use the same language (Tensorflow) as Akida does. ARM, Qualcomm are known entities that BRN is partnered with.

And we've barely scratched the surface - at the moment the AI capability of the Pixel is only used by Google specific applications, for example - live translation, speech recognition, song recognition and camera photo processing.

No third parties are writing apps for the Google Play store or Apple App store (to my knowledge) that take advantage of the AI processor. Part of that will be lack of ubiquity of the tech - if only Google is doing it with their Tensor Chip and Qualcomm isn't, it's not worth the effort. Google phones are only single figure percentage penetration of the overall market.

My assessment of the AI market is that the smartphone will become the hub for a lot of non-AI devices. Want to check your health/make diagnoses using a breath monitor? You won't necessarily have an Akida enabled device - you'll have one that pairs/connects with your smartphone and the work will be done on the phone. Same for a lot of other health devices for example there are bluetooth enabled blood sugar monitors that pair to your phone.

The advantage of doing it this way is that legacy "non-AI" devices will have access to Akida. For health devices that would be an advantage as new devices may need to be subject to a FDA type regulatory approval.

Run some AI smarts over your last road cycling session? Your watch/GPS handlebar unit won't need to have Akida in it, but the Garmin/Fitbit/Strava app in the phone will be able to take advantage of the AI processing capabilities onboard and do some extra analysis for you.

That's why BRN being involved with ARM and Qualcomm is such a big deal. I have confidence they are also involved with Samsung and Google too.