Qualcomm provided an overview last week of what they’re anticipating as we roll out products that use Qualcomm’s artificial intelligence (AI) technology, coupled with the rollout of 5G.
While you might think that 5G would drive centralized AI and computing capabilities, and you’d be partially correct, the fundamental advances will come when we couple 5G connectivity with AI at the edge. This outcome is because even with 5G, we won’t have the bandwidth to ship the massive amount of data generated by an equally massive number of Internet of Things (IoT) devices to a central resource for analysis. The related backhaul won’t carry the load.
Qualcomm and others are implementing AI at the edge and in the cloud strategy, where some initial analysis is done at the endpoint. Only the partially analyzed data is sent to the cloud for additional analysis.
This AI-5G capability anticipates a couple of technological advances we don’t have yet. One is a hybrid AI where the intelligence is distributed and centralized to optimize the available processing power and bandwidth. The other reminds me of something out of the movie “Minority Report,” where remote devices can recognize you and uniquely serve you based on what the system knows about you.
Hybrid AI is where some intelligence is at the endpoint and some is increasingly centralized in a cloud resource. Optimized both for the available bandwidth and latency and for the most efficient analysis of the data, the hybrid AI should perform like a centralized AI in terms of speed and outperform a localized endpoint, AI in terms of capabilities. Unlike distributed AI, which forms a peer relationship with others like AIs, hybrid AIs have more of a synergistic parent/child relationship between the cloud resources and the intelligent device.
A smart surveillance camera’s intelligence identifies the best picture for facial recognition. It only sends that frame to the centralized AI, which can now identify the tracked individual. Another implementation would be when using facial recognition to log in to a computer. The local AI decides you look like the correct user. However, before giving you access, it queries the remote AI for a second test based on recent breaches, gaining instruction on how to scan for that type of fake identity. For instance, taking a second shot, and if they are the same, assuming it is an infrared picture and not you who wants access, denying that access.
Auto Adapting Endpoints
In the movie “Minority Report,” Tom Cruise runs, uses biometrics, and every sign he runs by responds to him and provides an ad for something that he uniquely might want. When he changes his biometrics (by having his eyes replaced), the signs now think he is the eye donor. Even the e-paper newsstands put up news curated just for him (sadly, it is about his impending arrest). At the same time, the eye transplant option is a way in our future, while the ability for signs to use biometrics is possible today.
But not just signs, locks that automatically unlock when you approach or lock when they see someone approaching they don’t recognize, active pointers in stores to direct you to products on your shopping list using local signage, offices that not only auto-provision but direct you to the cubicle or office you are to use fo the day, and intelligent health care that, before the doctor even sees you, provides an even more accurate assessment of your immediate needs.
All this is coming from auto-adapting AI-based endpoints tied to centralized computing resources designed to alter the world around you to fit your unique needs better. That’s the most significant part of the coming change. Historically, we have to adapt to the environment. With hybrid AI, the environment will increasingly adapt to us.
Wrapping Up: The 5G-AI Future
Qualcomm talks about a fascinating future of hybrid AI devices that can identify who you are and automatically provide what you want. What about privacy?
Surveys have shown that as long as people get what they want when they want it, they don’t care about privacy. In a recent survey I’m currently reviewing, Amazon came out as the most trusted company, with Twitter and Facebook falling to the bottom of the list, and that is because Amazon uses the information they have on you to serve up better choices. Social media tends to sell that information to people who appear to want to do you harm, and thus the privacy concern is more about what is done with the data than with the capture of that data.
My surveys have shown a similar distinction. As long as the data is used to provide a better experience — Netflix serving up better choices, for instance — folks are okay with data capture. It is only when that data is used against them that they get upset. Now, what Qualcomm presented is a tool that can be used for good or ill, hoping governments get their acts together to assure the good and prevent ill.
I like the idea of a world that alters itself for my unique needs automatically. I can hardly wait!