Your AI is only as smart as the data that you give it - Unlocking the World’s most valuable data for the next generation of AI with Federated Computing | Eero Jyske

Introduction

All right, greetings from my behalf as well. I'm representing an organization called Airis, a startup company pan-European, originally from Germany but we have employees all over Europe. I'll be talking to you about how your AI, your models, are only as smart as the data you give it. Thank you for sticking around with the best track one where we also use AI to generate our images that we present. I've shamelessly also used Dolly mostly for my generation. You can see a bunch of typos here and there, and that's on purpose; I didn't go about trying to fix those typos that the AI is doing.

The Future - understanding the past and present

As Le was talking in her presentation, she took us from a bit of past, present to the future. There's no future unless you understand the past and present—where did we come from, where are we today, and then where are we actually heading?

Looking at the Past

I'm going to take a few moments to talk about first the past and then talk about maybe Tommy's world today, expose what you guys are doing in the future, but then I'll talk a bit about how do we make the AIs better going forward.

Le also mentioned Alan Turing, so I always tell my wife that the dream job I want to have one day is teaching the Turing machine and Turing theories in some institution. If anybody can help me with that, hook me up during lunch; I'd be very happy to take that opportunity.

When I went to the university myself, I had already done years of software development. I thought I was really clever and a fantastic software engineer. I went to the university, and the first class you get to take is the theory of computation. I took it, thought it was complete rubbish, and then I spent four years in university. The last two courses that I took were compilers and artificial intelligence, and then I realized, oh, Turing actually had some clever things to say. I read the class as the last thing I did in university, and I've been a big fan ever since.

This is Dolly's illustration of Turing as my superhero. Please make an image, and this is what it came up with.

I highly recommend a movie called 'Imitation Game' if you're not familiar with Alan Turing and the work that he's done. He's basically the creator of the computer as we know it today. Maybe quantum computing will change that a bit; I'm not going to dive into those details.

Moore's Law and Computing Growth

Another thing I want to emphasize is Moore's Law. Gordon Moore from Intel back in the '70s made a projection that the density of transistors will basically double every two years, and the cost will stay the same. In practice, that means that we're going to double the computing every two years, and the cost will stay the same.

A lot of the companies in Silicon Valley have successfully used this knowledge and built the empires that they have today, knowing and projecting what compute power will be available in the future and hence planning their products and strategies accordingly.

There is Huang's Law now; I don't think it's very official yet, but it's much more than doubling. Nvidia is currently racing at something like a thousandfold compute increase with the latest chips that they're doing. It's certainly increasing, but similarly, we can still predict what it's going to be, and companies should use this knowledge to understand what is going to be possible in the future—not be stuck with what you can do today but what is going to be the cost of doing what you dream of doing in the future.

Managing Software Complexity

In software engineering, what has happened during this time, from the '60s to today, is that systems have become bigger and more complicated. We can do more things, and a lot of the work that has gone into development in software engineering is to manage this complexity.

The Future of Data Collaboration

Now look at the collaborations, the way we set up to interact with multiple data centers. There's still a need for a third party like us to take part in facilitating and providing technology for this collaboration.

But in the future, obviously, technology is still needed, but there really shouldn't be a need for the middleman in between, making the data collaboration more day-to-day.

Whoever has valuable data can install this compute gateway type of device in their data center, publish and announce what kind of data they have, create examples, and give metadata to describe their datasets, and then monetize them to the interested parties.

This creates more of a data ecosystem where data custodians and data owners can advertise what they have, and similarly, the organizations that want to use this data can find the meaningful datasets and create these collaboration opportunities.

Again, going back to what I said about the direction, the value being in data will create opportunities for organizations that don't necessarily use the data themselves but can use what data they have to enable somebody else to do research on whatever they're doing.

Real-World Example: Protein Structure Prediction

One example of things that are already happening on this front is the Nobel Prize ceremonies. I didn't usually follow them, but this year, the chemistry prize was remarkably interesting because one of the use cases that we are helping empower is protein structure prediction, which is fundamental to a lot of medicine and drug development and other advances in healthcare.

If we can find ways to predict various protein structures, it helps us accelerate finding cures for lots of diseases.