If data is the new oil, then India is the world’s largest oil field. Today, India generates about 20 per cent of the world’s data, and yet has only 2 per cent of the world’s data centres to store and process it.
With the advent of the AI PC, data generation is only going to explode. That begs the question — are we as a nation ready to leverage on our most crucial resource? Or will this be yet another opportunity squandered?
We sat down with Gokul V Subramaniam, President and Vice President, Client Computing Group, Platform & Systems at Intel India, along with Santhosh Viswanathan, Vice President and Managing Director of Sales, Marketing and Communications Group (SMG) at Intel India, for a conversation around leveraging India’s data for its own benefits, and how India’s data security laws need to be designed to best leverage this resource. Edited excerpts:
Firstpost: Given how AI is easily accessible now, cybersecurity is turning out to be a massive and more pertinent issue than it ever has been. In such a scenario, what opportunities and challenges do AI PCs bring for end users?
Gokul Subramaniam: If you look at AI, there are three big things that you need to look to secure — the model, the data, and the generated data. And there’s a lot that can be done there.
On Lunar Lake, we’re bringing three IP blocks into our processors.
The first is a partner security engine, which is more for the OS layers. For example, if Microsoft has something at the OS level to authenticate, it can work very well with the hardware in a secure environment. So when you run models and algorithms, it helps in that one.
The second is a silicon security engine, which is primarily for authenticating that you need to have.
And then the third one is more along the lines of CSME, where you want to have these TPM things that you drive. So all these three are built in.
The one that is new is the partner security engine, where there is an ability for the OS to work with the system even at the hardware level. This way, even at the factory, where Intel makes these chips, or where where they are making a product, or the OS that is getting built, none of these three entities can see the user data. That’s the level of security the PSC brings.
We’re also always looking at AI-driven security aspects. And like I said, it’s the model, it’s the data that’s used in the model, and the generated data — all three need that level of protection.
Santhosh Viswanathan: I think the concept of AI PC also is aimed at ensuring that private data stays private. Because that’s what the user wants. They don’t want the data to be somewhere that they feel is not secure, they don’t want to share data that is private to them.
And if you can run LLMs, use your private data, and ensure that you get an output that is productive for you, that whole concept takes security to a whole new level.
Earlier it was about me giving developers and app owners data, and then keeping it somewhere secure and safe. Now as a user, I don’t need to give them the data. I can still process using my data, and figure out what I can do best with that.
I think that’s a very interesting angle to security, where it’s not just how we strengthen hardware-based security, it’s also how I ensure that the users are comfortable ensuring that their own data, which is in their own secure device, is something that they hold on to and not share.
FP: We are seeing a movement where chipmakers and OEM manufacturers are focusing more on on-device AI, as opposed to cloud-based AI. What led to this shift? Should users actually care whether their data is getting processed on the cloud, or is on the device itself?
Santhosh Viswanathan: There are two parts to this. Take India for example. We generate 20% of the world’s data but only have 2% of the servers, we simply don’t have enough server capacity. We have a capacity between 1 gigawatt to 3 gigawatts. We need to build capacity for sure, that in and of itself is not enough.
There are cities in the US that have more capacity than what we have in the country. We of course need to build enough infrastructure to be able to process our most valuable resource. I believe data is our most valuable resource because that is what will fuel our digital economy now. Not being able to store that data within the borders of this country is a huge risk.
If we are to harness our most valuable resource, we have two options. First, we go build and too at a particularly large scale. Second, and this is more important, we get started processing that data, the moment it is generated.
That is why we will have to build the capacity to compute and use compute in a more distributed manner instead of having it consolidated in these large big AI factories that focus only on the big problems.
We have to imagine AI in a way that is not limited singular area, to one big engine where it is being done on the cloud or a data centre — you have to imagine AI everywhere.
So the models get to be smaller, more practical. The bigger models will have to fuel smaller models.And you start to harness data and you start to build intelligence that changes everybody’s life.That form of scalable AI will impact our society much more. It’s like your Aadhaar or UPI — fabulous examples. It changes the way scale is.
For us to be an AI nation, the best way we can harness AI is to make AI everywhere. For that, it is important to focus on the end-user device as the starting source as much as we focus on building data centres.
Gokul Subramaniam: As to whether the end-user user cares whether it’s cloud or local, I think two important things come into play when people make such decisions — latency, and data security
Every PC will become an AI PC in the coming years, just like how we as a society moved from a feature phone to a smartphone. And I think that with AI PCs, that transition will take place much faster.
When that happens, we will need for the network to have the capacity to push everything onto the cloud, or, there will always be a need for things to get processed locally.
Moreover, there will always be uses people with use cases who would want to keep their data locally and securely on their own, rather than have it go to the cloud.
Let’s say you’re someone who creates video content — a teacher or a tutor, who’s teaching in one particular language but wants their video to be processed in five different languages in India, and wants to publish them all at the same time. Such a person would not want to keep sending his files to the cloud for each different language, and download each file separately and exhaust all their data cap.
They probably want to do it all on their PCs, locally, because there could be issues with latency, security, and finances as well. They will need to pay more for data, and perhaps even need higher bandwidth. All of these things come into play.
Santhosh Viswanathan: I think Pat (Gelsinger, Intel CEO) explained it really well, in his three laws (three Laws of Edge Computing). Law of Physics, which says you can’t just keep moving data up and down the cloud; Laws of Cost or Economics, where it’s just not financially feasible for you to keep sending data and processing it in a cloud-based engine; and finally, the Laws of the Land, wherein you don’t want to mess with the sovereignty of data where you send data far away from where it is generated. These are the laws that will govern edge computing.
FP: Based on this, what should India’s data policy be in that regard?
Santhosh Viswanathan: There are again, two aspects to this. First, we have to consider data as our most valuable resource — it is the new oil in many ways, only the thing with oil is that it is a finite resource, and is bound by the geography of the land.
Data on the other hand is dynamic. We have 1.4 billion people who are generating a wide array and a wide variety of data. We need to apply this in our industries, in our small and medium businesses, data is going to be everywhere. We’re going to be a much more data-rich society than anybody else in the world.
The second step is to protect your asset, and how do you protect your asset? You protect your assets by having the right infrastructure in an adequate capacity. You need digital infrastructure that can support public-private partnerships, who can then go build that infrastructure. More importantly, though, you start to look at innovative ways in which you can harness that data and ensure processing that data.
So, it’s a combination of both, as a society that we need to work on that will go help us go harness this information.
Link to article –
