eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Qualcomm officials continue to bring new capabilities to the company’s portfolio of mobile chip platforms as they look to take advantage of such emerging technologies as artificial intelligence and machine learning.
The company in February launched its AI Engine in a broad range of mobile platforms, such as its high-end Snapdragon 800 line of systems-on-a-chip (SoCs). It’s part of Qualcomm’s larger effort to bring such capabilities not only to the smartphones and tablets that the company has traditionally powered, but also to help expand the vendor’s reach into an array of adjacent markets, from PCs, notebooks and servers to network edge devices and automobiles.
In a July conference call with analysts and journalists, CEO Steve Mollenkopf said that Qualcomm’s “Snapdragon franchise has moved well beyond mobile devices … and is positioned to be a critical enabler in AI and machine learning at the edge.”
Now the company is bringing the third generation of its AI Engine to mainstream smartphones through its new Snapdragon 670 Arm-based SoCs. At a time when the workforce is becoming increasingly mobile and relying more on smartphones, tablets and other edge devices, AI is bringing not only improved performance, but also a real-time responsiveness, greater reliability and enhanced security, according to company officials.
The new mobile platform comes just as the use of mobile AI in the business world is starting to get underway. A growing number of smartphones – such as Apple’s iPhone X and Samsung’s Galaxy Note8 and Galaxy S9 and S9+ smarpthones it is Bixby voice assistant software—are rolling out AI capabilities, though initially the technologies are being used by consumers. Gartner analysts are predicting that by 2022, 80 percent of smartphones will have built-in AI features.
However, as AI is increasingly introduced into smartphones and other mobile devices, the business uses will expand, according to Bob O’Donnell, principal analyst with TECHnalysis Research. However, the infrastructure—such as the Snapdragon 670 platform—will need to be in place before the applications can take advantage of it, O’Donnell told eWEEK.
“AI in the business world is just starting to happen,” he said. “A lot of the opportunities will be around image recognition and natural language processing, but a lot of the applications will need to be custom-built to be effective.”
Essential to the AI Engine are the core hardware components that make up the SoC, including Qualcomm’s Kyro CPU, Adreno GPU and Hexagon digital signal processor (DSP). All are optimized to run AI applications quickly on the device, and the DSP includes sensor management capabilities that can enable algorithms used to enable AI-powered apps, they said. The third-generation AI Engine delivers 1.8 times the AI performance than the previous generation.
The AI Engine not only supports such company tools as the Qualcomm Neural Processing SDK and Hexagon Neural Network, but also other AI frameworks, including Caffe, Caffe2, TensorFlow, TensorFlowLite and Open Neural Network Exchange (ONNX).
“As AI grows more popular with mobile users, we’ve focused on developing processors that offer essential support for next-generation AI features, like smart camera settings,” Nitin Kumar, director of product management at Qualcomm, said in a post on the company blog.
In addition, the AI tasks can be performed on the device leveraging the AI Engine, so the device itself doesn’t have to be connected.
“Slowly but surely, all the stuff done on the cloud will be done locally,” O’Donnell said. “It won’t happen overnight, but it will happen.”
It’s also important that Qualcomm is bringing these AI capabilities to mainstream device, Avi Greengart, research director of consumer platforms and devices for GlobalData, told eWEEK in an email.
“The Qualcomm 670 adds AI capabilities previously only seen on Qualcomm’s more expensive processors,” Greengart wrote. “This means that companies can issue less expensive phones and use them for computer-vision tasks. Mostly this means better pictures—which can be important if you are using your phone as a document scanner, real estate, or insurance. With the right software, AI-powered computer vision can also be used for quality control and other specialized enterprise use cases.”
O’Donnell said having AI features in mainstream smartphones also means that more people will have access to them, which is good for users as well as for app developers.
“That means that all of a sudden, the number of phones available that can potentially leverage AI is larger,” he said.