- AI's true potential lies at the network edge and within vertical industries, where carriers can turn infrastructure into the backbone of the industrial revolution
- Telco humans are the antidote to AI’s ability to be both incredibly smart and shockingly stupid — all in one
- While Big Tech is charging blindly forward, carriers are developing responsible AI strategies, focused on building a sustainable and resilient digital future
Story time:
You (YOU!) are the head of R&D at a big corporation based in Mountain Alto, Calif., (we’ll call it Bigly Corp., in honor of our 45th and 47th President). You have invented a cool new technology called CI, or Clever Ish, that you think could change the world. You are asked to present your new idea to the board – no pressure, right?!
Huffy-puffy important Bigly Corp. exec:
“Tell us about your invention, lowly R&D minion whose actual name I am far too important to remember.”
You:
“I have invented a thinking machine that will make our company run faster, more efficiently and more profitably, largely by allowing you to fire lots of customer service agents and replace them with beep-boop robots.”
(A rapacious murmur of excitement ripples through the assembled board, including the representative of the private equity investor, hanging upside down from the ceiling in his cape.)
Bigly Corp. exec:
“Sounds great. Turn it on immediately without any further due diligence and let me know how many employees we can pink.”
(Cheers and huzzahs from the board; the PE rep smiles for the first time in a thousand years, revealing yellowed fangs stained with the blood of the working poor.)
You:
“Umm, well, we still need to work out some kinks.”
Bigly Corp. exec:
“What sort of kinks, lowly minion?”
You:
“Err. It struggles with some tasks.”
Bigly Corp. exec:
“Example?”
You:
“It can’t organize an Excel spreadsheet. Or do remedial math. And it doesn’t have any common sense. At all.”
Bigly Corp. exec:
“That describes most of the Bigly Corp. board. What else?”
You:
“It’s delusional. But not all the time.”
Bigly Corp. exec:
“Wait, what? Why?”
You:
“We don’t know. It won’t tell us.”
Bigly Corp. exec:
“I see. Anything else?”
You [shuffling uncomfortably]:
“Yes, if we ever try to turn it off, there’s a good chance that it will try to kill us.”
(There is a long pause. Bigly Corp. exec slowly looks around the table at the Bigly board members. He nods, once.)
Bigly Corp. exec:
“Well, I for one am not seeing a problem. I propose that we shut down our existing business entirely and invest a trillion dollars in deploying this new, untested CI technology ubiquitously across the world’s most essential systems. All in favor?”
(The motion passes unanimously; the private equity investor turns into a bat and flutters away to celebrate by draining the blood from a tiny baby.)
--End Scene--
Ok, so the above is fiction, but also not actually a million miles from the “let’s take the brakes off” approach to AI deployment now being endorsed by Big Tech and others with a literally vested interest in boosting AI stocks.
The enthusiasm for AI, especially in North America, overlooks the technology's innate defects, such as hallucinations and deception (Decepticons, roll out!). These have already precipitated serial problems in early deployments, from the hilarious (IBM’s chicken nugs fiasco) to the dystopian (racial bias and wrongful arrest in due to facial recognition errors).
Faced with these issues, and the consequences of getting AI wrong, the best way to deploy it would be to replicate the approach taken with all other new enterprise applications--where code is subjected to a multi-tiered level of scrutiny akin to what is required to get a new cancer drug approved, including development testing, system testing, user acceptance testing, and security audits.
But that isn’t happening with AI, for a simple reason: its inherent propensity to hallucinate means it is incapable of meeting the strict standards required for approval in enterprise network environments. And AI developers know that it can’t fail a test it doesn’t take.
That creates a significant problem: companies are deploying a technology that, when asked to predict the tenth digit of Pi, might well reply "a banana."
We wouldn’t allow this nonsense with any other tech, so why does AI get a pass?
Big Tech BS
It’s easy to blame Big Tech. It’s also appropriate to blame Big Tech and its “just do it” agenda.
When Sam Altman claims that OpenAI will achieve Artificial General Intelligence by 2025, or Google promises to switch its two-million-mile fiber network to hands-free Level 5 autonomous mode by Christmas and cure all disease within a decade, they are fueling a banquet of AI BS that investors and the U.S. media are eager to consume (like Ron Burgundy, they’ll believe anything Big Tech puts on their screens).
Deploying AI should happen for measured reasons, not as a way to help Big Tech companies quickly boost their share price (which will happen anyway, as a byproduct of getting AI right; it just takes longer than telling everyone to move fast and potentially break their businesses).
But a steady, attenuated timeline doesn’t cut it for companies like Google, AWS, Nvidia and Palantir. They want it all, and they want it now, or they’ll be very, very cross.
Fortunately for them, the U.S. government is happy to clear the way for them, for a price. This week, President Trump broke with constitutional convention (and possibly the law) by pressuring AMD and Nvidia to pay the administration for the right to export AI chips to China.
America is now at a point where one day, China is its sworn enemy, and AI is a vital advantage to be guarded at all costs. The next, it’s totally fine to sell those AI chips to China — so long as the administration gets a 15% slice of the cheddar.
The telco renaissance is here
Its ability to make embarrassing errors notwithstanding, AI still delivers many exciting benefits. But with Big Tech ignoring its defects, who can we look to for a solution? The answer might surprise you: telecom carriers. Their steady, cautious approach to AI deployment appears increasingly sensible compared to the reckless "damn the torpedoes" attitude often seen among Big Tech giants.
In conversations this week with AT&T and NTT Data, I noticed two themes. Both companies are intent on establishing the value of AI to their businesses and their customers rather than just diving in. And both of them are adamant that human beings will be in the loop in their telecom operations for the foreseeable future. Essentially, carriers are flipping the script on the concept of augmentation: instead of using AI to make humans smarter, they’re using humans to prevent AI from making a hallucinatory fool of itself while cratering essential services.
And in so doing, they solve an issue that has been around for four decades. AI hallucinations have been a problem since the first neural network bubble of the 1980s. (You didn’t know about that? Well, get ready to pick your jaw off the floor again when you learn that AI has been a thing since the 1950s. The big breakthrough that triggered the current AI boom mainly amounts to adding an annoying consumer chatbot to the front of an existing AI backend so everyone can have a play…)
The AI opportunity is a telco opportunity
Keeping humans in or on the loop on AI decision-making is the kind of measured approach that narcissist hyperscaler elites like to pour scorn on, but this pragmatism is what has kept carriers in business for decades.
And in the long run, carriers are in the sweet spot just as much as the Big Tech titans building data centers and chips; those are merely the AI engine room, the means of production. The real future lies in the “product” they enable: a new and astonishing digital industrial world.
While Big Tech and the investment community focus on the core — AI factories packed with GPUs, TPUs and NTUs — the truly exciting part of this revolution has already shifted to the network edge, and beyond, into the vertical industries connected through that boundary.
As companies and governments seek to reinvent their economies with AI, they can’t do so without carrier infrastructure — delivering the essential, ubiquitous, AI-native 5G connectivity for IoT applications.
But the telco role extends beyond simply providing bandwidth. Carriers have the opportunity to partner with systems integrators like Wipro, Siemens, Cisco and NTT Data — experts in the vertical applications running on OT networks within various industries — to connect these industries into the backbone of the new AI-driven industrial revolution.
Carriers who recognize this potential and embrace the opportunities at the network perimeter — and beyond — are poised to have a very successful 21st century indeed.
And the first shall be last
Unlike Big Tech companies like search engine company Google, online shopping outlet Amazon and for-profit-not-for-profit OpenAI, carriers understand that deploying AI isn’t a race; it’s a strategy. The goal isn’t to be first to make all humans redundant (or, if you’re death-cult meme-stock Palantir, to help kill or deport them).
Nor is it to be the first with a Level 5 autonomous network (Level 4 is the number to shoot for). If Google is truly dumb enough to put its fibre network into 100% human-free mode this December, carriers should step back and let it be the first to scale that particular mountain — and fall from it.
The telco operators I am talking to are developing mature strategies around AI, rather than trying to boost their stock price and egos by being first past the post.
Op-eds from industry experts, analysts or our editorial staff are opinion pieces that do not represent the opinions of Fierce Network.
Want more of this kind of thing? Subscribe here, and check out these fresh, related videos!