Jason Zander, executive vice president of strategic missions and technologies at Microsoft, is responsible for bringing the Microsoft Azure cloud computing service to the heights of success it currently registers. A few days ago he visited the Mobile World Congress in Barcelona, ??where he presented to telecommunications operators some new services based on AI, such as the detection of fraudulent calls.
Until recently we talked a lot about 5G, but now it’s all AI, how does that change?
First I will give you a fact: at this moment the world GDP is around 100 trillion (trillions in English). Our estimates are that generative AI can add $7 to $10 trillion in new global GDP, which is pretty incredible from a growth perspective. We believe that if operators are able to embrace AI, they can modernize and monetize their networks through a combination of cloud software and generative AI itself.
How does this apply?
The cool thing about generative AI is that it can also reason about network topology. That is, the design and the systems in place. In fact, it can reason about all the current and active cybersecurity issues in the world, all the vendor documentation, and that sort of thing. So, aside from looking at data flows, I can understand my entire network. Combine that with Azure Operator Nexus, which allows me to manage my data network, and suddenly you can do things in days that used to take weeks or even months to do.
In Barcelona it has presented the detection of fraudulent calls. That is?
It turns out that AI can actually listen in on a conversation, detect fraud, and actually break into the call and say, “I’m your operator. We have detected that this may be a fraudulent call. Allow me to explain it to him in case he wants to continue the call”, because the user must be able to decide for privacy purposes. But it is a great example of using artificial intelligence to protect consumers. And also from the perspective of the operator. It is also a new source of income, because it could be a subscription service that the user hires to help them protect themselves.
How can AI detect these fraud calls?
We won’t reveal all the keys. Hackers will try to avoid this, but if you think about what happens in a scam call, they generally try to create a sense of urgency. AI detects these kinds of patterns. These are the warning signs of fraud. This way he can interrupt the call and put it on hold, and the subscriber is given the opportunity to think about whether it is really someone he knows. The operator can even decide whether to add a human in the loop to help.
But this will require more data security and privacy.
Indeed. You have to implement another layer of security. We want to make sure that we respect privacy at all times. It would be a service that the subscriber would have to contract and we would certainly never listen to the calls unless it was a service that the user wanted. BT is the first to try it in the UK. I have two elderly parents and I would subscribe to this to help protect them.
Are there areas of telecommunications where AI is better?
Yes. For example, in the operation of the network itself. In this case there is a bit of machine learning that is being used in many networks today. Got something misconfigured? Do you have a security vulnerability? Do you have something that is out of date? In all these kinds of things.
AI seems to be solving problems we didn’t know how to solve for years.
At Microsoft we have Copilot, and we call it that on purpose, not autopilot, because in this case, the machine would solve it for the person. We don’t think that should be the case, we know that sometimes it will come close and it might not be exactly right. And then we keep thinking that the professional, whether it’s a doctor, a lawyer, a network operator or a software developer, will get help from AI, but that’s not meant to be perfect, is it? That’s why we want humans to get involved.
Can you predict what this might look like five or ten years from now?
What I can say is that the models, the bigger the supercomputers, the more data we use to train them. And as we fine-tune the algorithms, they become smarter, and we don’t see the end of this trend yet. It’s the reason the industry is deploying so much computing power and so many GPUs to do training, so it just keeps getting better and better and better. And I think the kind of general-purpose AI, which is what this generative AI is providing, is applicable to everything. It is enough to talk to some of our colleagues to realize that it is real and that it is already here. The funny thing is, this wasn’t there a year ago.
Is Microsoft the leading company in the use of AI in the world?
Well, I would rather say that we have a very strong commitment to AI. It’s been seen with the multi-million dollar investments we’ve made in various areas, and I’d like to point to our product roadmap as well. The fact is that we have launched Microsoft Copilot for multiple different products, for security, for software development for research, also an office and productivity, etc. These are things we’ve been working on for a long time and they’re here and people can use them today. So from that perspective, I feel very good about the core technology that we have invested in for years. And I feel really good about the products we’re already launching. And we have only just begun.
