The combination of 5G networks with artificial intelligence may help machines better understand language in context and feed their ability to learn.
Artificial intelligence (AI) can make devices do amazing and sometimes numbskull things. It’s still common for voice-activated digital assistants to grasp the meaning of a spoken command, but misconstrue its true intent because there’s no context.
That’s what happens when services like Alexa, Siri and Google Assistant misinterpret homonyms and homophones. Mistaking rose (flower) for rose (lifted) or “where” for “wear” can often lead to hilarious responses.
Understanding context is critical to human interaction with the world. It’s equally essential for AI-powered services, according to Bob Rogers, chief data scientist for analytics and AI in Intel’s Data Center Group.
He sees 5G, the next generation wireless technology slated to arrive in the next year or two, poised to improve context awareness in AI, making these services smarter, more useful and prevalent in people’s lives.
“The ability to understand context is going to change how we communicate with our computers,” said Rogers. “Personal assistants will be able to understand us better because they will understand the context of our questions and to give us useful information we seek.”
Depending on how the word “tablet” is used, for example, it could refer to a handheld computer, a writing pad, a pill or even a large, flat stone like the one containing the Ten Commandments.
“But if a human being is in a doctor’s office and is told to swallow a tablet, that person immediately knows they should swallow a pill not a computing device,” he said.
Faster, Smarter Connection to More Data
5G connectivity has more bandwidth and lower latency than current wireless networks, so it will transmit more data faster than current 4G technologies. For example, LTE tops out at 1 gigabit per second while 5G is built to deliver 10 gigabits of data per second.
The ability to access additional information quicker through 5G networks will help AI-enabled devices understand their environment and the context in which they operate, said Rogers. This will supercharge AI-powered services, making them more reliable in a broad range of situations.
More bandwidth may help digital assistants like Alexa learn to distinguish between voices in a household, so online services will only take orders from mom or dad. This could keep children from breaking the bank by inadvertently ordering items like dollhouses or cookies.
Computers have always been good at handling information that is structured, where the data is provided in a consistent and standard way, such as numbers and equations. But humans, and most of the world, deal with unstructured information from a hodgepodge of audio, video, text and numbers.
“AI is just beginning to be able to handle unstructured information,” said Rogers. He pointed to social media platforms that now use AI to enable facial recognition and photo tagging.
The emerging Internet of Things (IoT), including smart cities and autonomous vehicles, will need to control vast amounts of unstructured data very rapidly. As the world is increasingly populated by things that sense their surroundings and can communicate with each other, Rogers said the combination of AI running through 5G networks will be crucial and open new opportunities.
Faster Data and Processing Powers Instant Understanding
Traditionally, most AI applications have resided in the cloud. Rogers sees a future where more data is processed in cloud computers located near the device (for example, a smartphone or security camera). He said very low data transfer latency will be vital to many emerging technologies, including autonomous cars, smart cities and healthcare services.
With self-driving cars, for instance, the decision of braking or accelerating must occur with near-zero latency, according to a recent report by McKinsey. Edge computing will also emerge as the favored choice for applications where privacy issues and data bandwidth are paramount, such as AI-enabled CT scan diagnostics.
Keeping the data and processing on or near the device can lead to higher performance and tighter data security, said Geoff Blaber, vice president of research at CCS Insight.
Blaber said that 5G’s low latency will allow more computing of data produced by IoT devices at the edge of the network before being sent to the cloud. For example, AI uses computing power on a smartphone to tag photos instead of constantly communicating back to the cloud and creating lag in the user’s experience.
“Users and companies don’t necessarily want their data traveling back and forth to the cloud,” he said. “5G is going to be a very substantial enabler of increased capability at the network edge and a reduced reliance on the cloud as we see it today.”
Edge computing could also mean a device will be able to understand more about the context in which it operates, which will ultimately make AI an essential part of everyday applications.
“AI is really the practice of doing activities with computers that were previously only within the realm of humans,” said Rogers. “But to do that well, AI needs to understand a particular situation or piece of information, then put it into context before it responds or takes action.”
In fact, 5G’s low latency, high speed and edge computing capability could ultimately enable devices to communicate directly and work together, something called “swarm computing.”
For example, individual drones surveying a bridge could share digital images of one side of the bridge in real time. Then computers could use AI to stitch those images together with related archived pictures to build a 3D image of the bridge, said Rogers.
Devices Become Teachable
Ultimately, the combination of 5G and AI will make devices teachable, said Rogers.
“With the next generation of machine-learning algorithms, AI systems will use new information to incrementally build up understanding of their environments,” said Rogers.
It will empower digital personal assistants and other devices to learn their owners’ environment and preferences at any given moment, and through time. That means they’ll come to understand that a “tablet” in a doctor’s office is different from a “tablet” computer.
“Wrong answers and the feedback from users is what improves the algorithm,” said Rogers.
He said fast, reliable 5G data connections will allow AI-enabled devices to quickly learn from their mistakes and understand situations in order to get the job done right.