With this enormous opportunity, I believe comes enormous responsibility.
By now we should all be used to the apocalyptic and dystopian paranoia about AI stealing jobs against a backdrop of the rise of the robots. There isn’t a week that goes by without some learned body or vested interest pumping out reports warning of the shape of things to come.
But yesterday it was the turn of the CEO of tech firm that’s betting the farm (in part at least) on AI to deliver some cautionary words about where society could head if the correct checks and balances are not put in place.
Satya Nadella, CEO of Microsoft, used his keynote address to the Build developers conference to affirm his credentials as a believer in the potential of technology, but added a spin to the usual pitch by staking a claim to pragmatic realism about the downsides:
I’m an unrepentant tech optimist, there’s no question of that. But also I’m grounded, though, that there are unintended consequences of technology.
With that, he reached for the cliches more familiar from the work of over-excited PR flacks and cited 1984 and Brave New World as examples of what could happen if society’s not careful:
I do believe that it is up to us to ensure that some of the more dystopian scenarios don’t come true. I mean, if you think about it, what Orwell prophesied in 1984, where technology was being used to monitor, control, dictate or what Huxley imagined we may do by just distracting ourselves without any meaning or purpose, neither of these futures is something that we want.
But of course with AI tech to sell to the masses, Nadella was only ever going to default to an ultimately upbeat conclusion. With that in mind, he proposed a number of specific actions points to avoid future problems:
The first one is that we should empower people with technology. That’s got to be the front-and-center consideration for everything that we do going forward. People have to be in the loop. Let’s amplify their capability, amplify their ingenuity.
Second, let us use technology to bring more empowerment to more people. When you have these amazing advances in computer vision or speech or text understanding, let us use that to bring more people to be able to use technology and to participate economically in our society. So this inclusive design, ultimately, can be an instrument for inclusive growth.
Lastly, I think it’s up to us as the world becomes more technology driven, digital technology is part and parcel of every aspect, building trust in technology is crucial. And I think it starts with us taking accountability, taking accountability for the algorithms we create, the experiences that we create, and ensuring that there is more trust in technology with each day.
Nadella also touched on the speed of innovation and change, noting that the themes at this year’s Build have moved on since last year’s gathering to take on more and more of an AI flavour:
Last year, we talked a lot about conversations as a platform, and this year, you will see in demo after demo, people using conversations as a platform. People building agents, personal assistants that span devices. Skills for those agents, bots, natural user interfaces for these multi-device applications, whether it be text, speech, or even vision and gestures with mixed reality.
AI is everywhere. The ability to reason over large amounts of data, create intelligence, distribute it. Even the micro services, workflows, advanced analytics that people are building in the cloud are all pointing to what I think is a fundamental change in the paradigm of the apps that we are building, a change in the world view that we have. We’re moving from what is today’s mobile-first, cloud-first world to a new world that is going to be made up of an intelligent cloud and an intelligent edge.
From the point of view of the developer audience he was addressing, Nadella said there are learnings its members need to take on board in terms of the fundamental characteristics of a new application pattern and world view:
First, is the user experience is getting distributed across devices. It’s no longer just mobile first, in other words, it’s not about one device, and app model for one device. The user experience itself is going to span all of your devices.
It’s easiest to understand it when you think about your personal digital assistant. Your personal data assistant, by definition, is going to be available on all your devices. And, in fact, as you move between devices, it’s going to be there, helping you get your tasks done.
That multi-device experience is what now needs platform capability, you need new abstractions, not to just a single piece of hardware, but all of the devices that help you build out your application.
The second major consideration is AI itself, he went on:
We know the currency of this platform shift is all about data. We [have] talked about the car, right? 100 gigabytes of data per second for an autonomous car or connected car. When you have that type of data being generated at the edge, data has gravity, computational power will move to it.
That means the AI you create is going to be, by definition, more distributed. You will, in fact, already see these cases where people are doing a lot of the training in GPUs on the cloud, but are deploying the models on the edge. But we’re reaching a point where you can’t rendezvous all the data for training in the cloud, you actually will have to train and do your entrances in the edge, so you absolutely need a new set of abstractions for even AI in terms of how you learn and you train and you infer that span both the edge and the cloud.
As data and AI get distributed, there needs to be a fundamental change in the inner and the outer loop of what developers do, he added:
Continuous integration, continuous deployment is going to change. And that’s where things like micro services, containers are playing a massive role in the outer loop. But one of the things that I think is going to completely change how we think about logic is serverless.
In a world where things get more distributed, event driven, you want to be able to write logic that reacts to these events, is not static. That means you don’t have a piece of logic just being bound and running on one virtual machine in one location, but it’s something that can actually be mobile.
I mean, think about even these AI models. One of the big challenges is going to be model drift. So you want to be able to tackle that. So serverless computation is going to fundamentally not only change the economics of what is back-end computing, but it’s going to be the core of the future of distributed computing.
Nadella is always at his most eloquent when talking to the developer community. This was no exception. Inevitably the mainstream media has picked up today on the ‘Microsoft boss warns about AI’ angle, but this was in fact a solid rationalisation that raised some interesting discussion points for the rest of this year’s Build conference.
Image credit - Build17