Living with AI

Max Welling, Research chair in Machine Learning at the University of Amsterdam and VP technologies Qualcomm Netherlands, talks about reschooling, automation, employment, jobs, cloud computing, edge computing, servers, connection, delay, energy, battery, internet of things.

Share Discuss

JOBS AND AUTOMATION

In terms of the job market, many people have claimed that AI will displace a lot of workers and in specific areas this might happen, for instance, truck drivers: if the self-driving car and self-driving trucks become reality, it could happen that a certain generation of truck drivers loses their job and find it difficult to re-school into a new job. But, on average, in general I think that even though the labor market will change, certain jobs will go, but also other jobs will be created around the opportunities that are provided by AI. The challenge maybe is to identify those new opportunities and to re-school people into those new jobs and so there's a big responsibility there for universities and schools to make sure that people can quickly re-school into these new opportunities.

 

PRIVACY

I believe we can create a world where privacy will be maintained, even though we will develop very powerful machine learning and AI with them. But we will have to change things, currently our data get absorbed into big companies and we don't think much about what happens after that. If we would change the system, where we own our own data and we have control over our own data, then you can imagine systems that learn off of that data in a way that is even privacy preserving.

 

In other words, if an algorithm comes to my data and says “I would like to use your data to train a little bit of my model” then you could specify what level of privacy you require for that update, and then it will only look at sort of averages. It will not actually update the model in such a way that later somebody else can recover a set of sensitive data on that you provided for that learning. I strongly believe that in the future we will see the emergence of new algorithms, which are perfectly privacy preserving, but still can learn from all the data that is around.

 

EDGE VS CLOUD

There’s two ways that computation can be done, especially to power AI algorithms. One is you compute and train in the cloud, which is a bunch of servers somewhere, in some place, and so you would have to send the request for processing to those servers and there's clearly delays involved, also it might be an unreliable connection. But also, you might not want to share your data with the service, so there's a bunch of disadvantages. The other thing is that you can actually do the computation on your device locally, so that means that you don't suffer from these delays or latencies, and that you can maintain privacy.

 

Of course, in order to be able to do that, we will have to be able to build powerful enough algorithms that they can run on the device, also in such a way that they you don't use a huge amount of energy, otherwise your battery is drained very quickly. But, I do believe that these algorithms are possible, and we are developing these algorithms. In the future, I think more and more of the computation will actually happen on the edge, on the devices that we have. In terms of the impact that it will have on the internet of things, of course we will put more and more devices everywhere, in our cars, in our homes, in our offices and we carry them on our bodies, and so if we want to have AI applications running on those devices, we will have to be able to do the computation locally and on the device.

 

AUTONOMOUS DRIVING

In terms of automated driving, in the beginning when the systems are not perfect and that the media attention is very much focused on it, every accident will probably be under a magnifying glass. But if you look at the actual statistics we’ll probably find that it’s actually a lot safer to have automated driving, then have human driving. It's a matter of trust, so will we, as a society, start to trust these cars? I believe that we will, because at some point in the beginning it might feel uneasy but, at some point, we’ll find, we will have the experience that they actually drive very safely. Part with the actual statistics, that the number of accidents will go down, not go up. We will probably start to trust these systems, like sometimes we also trust planes that they will land for us, although there is still somebody in the cockpit, I do believe that in the future we do not need somebody behind the steering wheel anymore.