Spoiler alert: This article begins with AI healthcare and ends with broadband network maintenance and the NTCA 30-Day AI Challenge. Intrigued? Read on.
I was at a friend's gym this week when he pulled a small device off the shelf and said, "Watch this." It was a wireless ultrasound transducer, about the size of an iPhone. Always cautious, I qualified, "I'm not pregnant," as he reached for a bottle of gel, which he then (to my relief) squeezed on his elbow. He ran the probe back and forth, captured several images, and by voice command transferred them to a tablet before uploading them to ChatGPT. "Tell me what you see," he asked. ChatGPT "thought" for a moment and replied that it was an elbow or knee before describing the condition of the joint, tissues, and tendons. My friend scratched his head and said, "Well, that's about 50% right." He then transmitted the image to OpenEvidence, an AI tool accessible primarily to users with medical licenses. OpenEvidence made the right call and provided a detailed, accurate analysis. We then fed the app stats from a fictitious patient - age, height, weight, symptoms and preexisting conditions - and asked it to craft a prior authorization letter for a specified medication. Within seconds, the letter appeared, complete with citations and prompts for the "requesting physician" to fine-tune certain details. "Something like this would take me at least an hour," my friend said. "Now I get it in seconds, and after a quick review, it's good to go."
And there we are. AI healthcare tools are becoming increasingly accurate. Physicians participating in a recent study preferred AI answers to those of other physicians in eight of nine clinical areas. And while doctors still preferred consulting with specialists over AI, both specialists and general practitioners rated AI answers as safe as human physician answers. Diagnostic abilities of tested AI bots were found to be almost as good as doctors', though humans still led on triage. But AI is getting better. Performance accuracy among several tested AI medical models improved from about 33% to nearly 90% between December 2020 and May 2024. Applied to rural settings that lack access to specialists, the potential is tremendous (see our prior discussions here, pages 7 and 8).
Now let's break this down for the telecom space. Remember, AI is all about patterns - words, languages, data, images. When we talk about healthcare AI reading data to create diagnoses and treatment plans, or agricultural AI supporting "see and spray" systems, or large language models (LLMs) enabling advanced customer service chatbots - those same tools that spot and understand anomalies can be applied to network management.
For example, vast yet intricate data traffic patterns can be analyzed to identify congestion points and spot early signals of security anomalies like DDoS attacks. Or use AI to monitor Rx/Tx (receive and transmit) levels and flag subtle changes that could signal brewing troubles or an excessive bend (it all comes back to elbows). Or field technicians can be armed with image-based diagnostic and remediation tools. Do we still need field technicians? Of course. We're not talking about payrolled robots. We're talking about giving people next-generation tools. Try it at home: take a picture of that leaky faucet, ask an LLM what part you need to fix it and then ask for detailed instructions. You'll be the star of the house in no time.
But you can do that on the big stage too, with the NTCA 30-Day AI Challenge. This contest has two divisions (Newbies and Denizens). It's about test-driving tools and learning not only how they can increase efficiency but how you can navigate them to get useful and accurate information. All in an engaging, judgment free zone. A test drive on a private track.
As a starter idea, ask an LLM to design an image-based protocol to help technicians diagnose problems in the field. Or a tool that can record and transcribe customer inquiries and help CSRs achieve FCR (first call resolution). And then share your stories for spot on the big stage.
It's pregnant with possibilities.