Home » AI World Society » AI World Society Summit » AI World Society Summit » TRANSCRIPT OF Professor Alex Sandy’S SPEECH AT AIWS-G7 SUMMIT INITIATIVE


OK. Hello. My name’s Alex Perry Well I’m a professor at M.I.T.. And I run a group. Called connection science which actually develops techniques and. Open source software. For helping countries. And. Companies. Deal with A.I. in a way that is both effective. And efficient. But also ethical. And. The key idea is. That in this new world we’re entering as you have been everywhere. Most of the day was in private hands in the hands of. Companies. By yet you need that data to be able. For it. To be able to run government to be able to be efficient about social systems. Civic systems. How are you going to do that in a way that’s trustworthy unbiased fair. And that the people understand. What’s happening. And. So. I’ve developed a method that I call. Open algorithms. That allows you to take data from private entities like companies. And combine it. In a safe way. To produce. Data. About. Entire countries or other large structures large. International companies etc.. In a way that’s safe. Auditable. And. Understandable. So. The drivers of this are things like. Security. If you put all the data in one spot it’s sure to get stolen. An enormous risk. It’s. Inertia. Different people own the data. And. With privacy concerns. Individuals. Own the data in a certain sense. They should be able to know what’s happening. They should have control over data. About and. How do you put this together. The big. Idea is that instead of putting data on one place you keep data where it’s originally collected. This is you. Federated data. Rather than concentrated. So the people that have. Ownership. Rights over data. Continue to hold on to the data. But they agree. To. Algorithmically answer questions. So. The computer that data is on is also willing to answer questions. For the public good. Or for. The Joint Group. Those questions are called. Algorithms. And. In our system we call these open algorithms because. They’re things that are. Legally agreed to beforehand. So instead of going to someone to say can you give me your data and you say. Well here’s an algorithm I’ll show it to you. Can you run this on their data and give me the answer. Legally and organizationally that’s a much easier thing to do. And in fact. Small countries like Estonia have been doing it for a long time. And big companies. Like. AT&T. Do it as a matter of self-protection. To make sure that they’re not breaking the law. Well it sounds complicated to the human. Ear. I’m a computer point of view. Reading the data where it’s collected and answering questions. Is. In many cases more efficient than moving the data to one place and answering the questions there. The difference is is that it’s a lot safer to leave the data wherever it’s collected. Military people figured this out a long time ago it started building a castle with nodes. Which. In today’s computer Oracle firewalls and they work to defense and debt. That’s what we do with. Data. When you do that all of a sudden you can keep track of what questions are being asked and what data. And what people who collected the data and the people who already. Have the ability to monitor if the. Questions are ones that they agree to. And. So you can record what’s happening in the system is orderly. So if you have a question about bias or fairness. You can go. And. Answer that question because now you have a record. Of what was done with the data. And who did it. So for instance. In the country of Columbia we were able to look at. Know. Poverty programs. And discover that there were almost a million people. Who were getting benefits that shouldn’t. And a million people that weren’t getting benefits. Sure. And that comes with the ability to. Audit. All. The decisions. Which are going to human ears sounds complicated but actually. Computer point of view. It’s just a dashboard that keeps track of what’s happened. So. That’s the sort of big view is that instead of having. One centralized control one centralized repository. You have a federation. Of different. Players. And their interests that agree to answer certain questions for certain functions. And you ordered them. So. As I said there are some countries like Estonia that are doing more. Recently. Europe. Agreed. Eurostat the official data. Organizations of all. EU countries. Adopted this sort of framework. And there are several other countries that we’re working with. Israel and Australia. Others. That are also. Putting pilots in this to be able to explore. How they can get better insights about their country. And had better policies. By using this public and private data together. In an. Open. Honorable way. So. That’s the key thing. We’re. Deploying these things in different place parts of the world. We’d be happy to help you if you’re interested. People are typically interested in things like. Social programs being more efficient being able to have greater income from tourists or from. Innovation. Or new sorts of civic systems better transportation and public health. And we help build those. Things for people. We don’t know turnkey. Solutions what we do. Is we build. Prototypes. That are then. Specialized to be operational for your particular situation. And if you want to know more about this I refer you to for instance. The keynote again for the EU presidency. Or. Other sorts of talks like that that. Will be making available. To you. And we have a book called trust data. Which describes the techniques that includes the. Piece that the Obama White House asked us to do for them. The piece that the UN secretary general has. A piece for The World Economic Forum. Describing both the policy the legal and the technical aspects. Interestingly the Chinese central government just translated this into China on. Chinese and published it through the. Chinese. Central and economic press. So you might be interested in. The. Problem. So that’s the top level story here. I hope you’re interested we’d be happy to talk to you and. And work with you. Thank you