How telecom can blend intelligence and connectivity with AI

Artificial intelligence seems new, trendy, disruptive, threatening, world-changing - a great leap into the unknown. But will people see it that way in, say, 200 years? Or will it be a step grounded in history, limitations and practical applications?

Ericsson blog ai telecom connecitivity

Panelists at “Artificial Intelligence: Underpinnings of a disruptive wave”, an event hosted by Mobile Monday at the Google Launchpad space in San Francisco, put AI into this kind of framework, kicking off with the moderator Chetan Sharma, who took the audience back all the way to the beginning of the 19th century to the roots of thinking about artificial intelligence (which can be taken even farther back to myth and Greek philosophy).

Explore how AI can help networks evolve from connectivity to predictivity

Then Sharma pulled us into the present, talking about how many AI startups were funded last year, massive investments into enterprise AI and the societal effects of automation, particularly job losses.

AI startups

Putting AI in context

The participants tended to downplay AI hype and focus on the practical everyday uses for it. For instance, Barak Turovsky, Head of Products – Google Translate and Machine Intelligence, said he saw two primary uses for AI - in cases where you have too much unstructured data and in areas where you’ve reached a plateau with current approaches and need a jump (such as translation).

He added later that it would just be too hard for the current generation of neural networks to make all the decisions in a certain field.  He saw AI as more “assistive,” using the example of a radiologist who is accurate 90-95 percent of the time.

“AI can say:  ‘Look at these two or three exceptions. Take a closer look,’” he said.

David Gerster, VP – Data Science, BigML, talked about the evolution of customer service, as people try to figure out if they’re talking to a bot or not, and how they might, in the end, prefer the bot to the real human.  While Tsvi Achler, CEO, Optimizing Mind discussed aspects of AI that frighten many people - it’s lack of explainability, for instance - as a real flaw.  If AI can’t explain itself properly, and if it can’t update its process in real-time like a human can, what use can it be for critical applications?

Because if Siri on your iPhone or Alexa from Amazon make a mistake, we can shrug and laugh.  You can’t do that in, say, a telecom network.

How can AI help mission critical networks?

Diomedes Kastanis, the Head of Innovation of the Ericsson Innovation Office, stood out on the panel because he has distinctly different focus than your typical AI company.  He’s concerned not with applications running over the network but with the actual network.  What role does AI play in network evolution?

Kastanis laid out the four steps as in a VentureBeat article he wrote earlier this year, with a focus on performance tuning and network management:

  1. Traditional networks
  2. Predictive networks
  3. Prescriptive networks
  4. Self-healing networks

“We try as an industry to blend intelligence with creativity,” he said at the event. “Having smarter networks will benefit things they’re connected to. If you make networks more predictable, the overall impact will be better for society.”

Heterogenous data and acting in real-time

Networks face two significant obstacles, Kastanis said after the meeting.  They have to deal with heterogenous data coming from all over the globe and from a huge variety of systems, vendors and network types.  And then they must be able to act in real-time.

“We see different traffic patterns, different geographies,” he said during the event.  “It’s not easy to just say that something that happened in China or Singapore will happen in the US.”

He added that even the same type of installation - he used Voice over LTE (VoLTE) as an example - could react differently to different traffic patterns.  So there could be similar problems - dropped calls - but with totally different root causes.

Read the new report from the AT&T Foundry, Ericsson and RocketSpace:  The Future of AI in Consumer Experience

How AI will help 5G networks

AI has the potential to predict and analyze issues faster than humans ever could.  He pointed particularly to 5G, a potentially transformative platform for industries, the Internet of Things, and vehicle infrastructure.

“5G will be more predictable, open, adaptable,” he said. “We will use AI and other techniques to make it more open as a network - enabling connectivity to predictivity.”

You can check out the entire panel here:

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.