Qualcomm bets on AI to take Google apples and chips
But this year,
Google has thrown a hat on the ring by debuting the indigenous tensor chip to
provide a Pixel 6 platform to showcase the company's vision for smartphones.
The massive
encouragement by this technology giant into the chip game means that Qualcomm,
so leading players in the smartphone industry, will face more competition than
before. So, this year, Chipmaker is very dependent on AI to provide a new
processor, Snapdragon 8 Gen 1, a little more personality.
The company
introduced the Qualcomm AI-7th Gen Computing Engine, which was four times
faster than its predecessor.
The chipmaker
claims that the new unit is stronger because it will double the shared memory
and tensor accelerator doubled for AI related operations. Plus, battery
efficiency 2.8 times better than the Snapdragon 888 flagship processor this
year.
And besides,
there are new 3rd-sensing units that can activate very low-powered voice and
connectivity experiences.
So now after we
see the new AI Chip specification, let's see what it means for telephone
makers, developers, and consumers.
What can the new AI unit do?
VINESH Sukumar,
Head of Product Management AI / ML in Qualcomm, said the new AI unit can handle
more workloads easily. The company developed a demonstration with ArcSoft video
imaging companies, where cellphone prototypes based on Snapdragon 8 Gen 1 can
simultaneously detect face and apply the effects of bokeh's face.
Sukumar said
that new technology allows software to detect more than 20 points on the face,
so that the application maker can apply the effect even when there is a bad
perspective or small display field.
New Hub senses
allowing chips to always be on camera experience for facial detection and object
while consuming low power.
There is also a
new 'always safe' module in a hub sensing that allows the application to make
financial transactions in a safer way.
Why does Qualcomm partner with companies for various experiences?
Until now,
Qualcomm chips usually provide telephone makers with canvas to draw their own
paintings. While it was still possible, this year, the company partnered with
many companies to provide a different AI experience a little more talent.
First, for an
enhanced-camera experience, including charming black and white shots and the
effects of bokeh, Qualcomm partnered with the German imaging company Leica.
Previously working with producers such as Huawei and Xiaomi. But this is a rare
opportunity when a chipmaker works with the camera maker exclusively.
Sukumar said
Leica's camera maker was the main partner in developing AI to capture
high-quality and solid images with great spatial details. He said that this
technology would increase the dynamic range and clarity in the image.
Qualcomm also
partnered with the Asante Health Organization to show off applications that
record your voice for 30 seconds and detect stress levels, asthma conditions,
and even Covid-19.
This is not
fully novel: Israel vocalist startup has achieved similar achievements with the
application, but in the case of Qualcomm, all processing is done locally to
maintain the privacy of users to remain intact.
One of the most
important partnerships is to embrace the face, open-source startup that works
on a language-based machine-learning model. Qualcomm Demoed Smart Assistant
that uses sentiment analysis to classify the most important notifications for
you. Plus, you can see a summary of notifications based on their priorities.
The assistant
can also answer questions like "Where is my dinner meeting?" And
"When do I have to go?"
All of this
reminds me of Google Assistant. But Vice President and Head of the AI
Qualcomm software, Jeff Gehlhaar, told me that this partnership would bring a
portfolio to use a broader use case. I would be very interested in seeing the
assistant who had a better ability than Google Assistant, so just take it to
Qualcomm.
What's in it for the developer?
Qualcomm said
this year, it introduced support for mixed precision instructions in the SDK,
so the developer can make the AI model they run faster when using less
memory.
However, the
biggest change is to introduce the search for nerve architecture (NAS) for
Qualcomm chips. NAS allows data scientists to optimize models for certain
hardware without training them manually. They can also place constraints on the
size of the model or other parameters based on use cases.
Chipmaker
partnered with Google Cloud to provide the backbone for this NAS offer.
Qualcomm will use the Google Vertex AI solution for managed machine learning,
so the developer can train their models easily.
This partnership
will allow developers to set up their models not only for cellular but for XR,
IOT, and automotive Qualcomm solutions as well.
Battle with Apple, Samsung, and Google
Apple always has
its own chip on its iPhone, while Samsung uses a mixture of custom exons
processors and Qualcomm Snapdragon processors for their devices.
This year,
Google waste Qualcomm for the first time, so that competition in the chip
market must heat up.
The giant chip
based in the US needs to show its processor that is capable as Apple, Samsung,
or Google, if not better. And for that, it puts the responsibility for his AI
meat.
It would be
extraordinary to see companies like Oneplus, Samsung, or Xiaomi using
Qualcomm's new equipment to issue AI experiences that we have never seen on an
iPhone or pixel. Game on!
Post a Comment