Meet VeeMe: The Virtual Agent Programmed to Think Like Me
If you think VeeMe sounds creepy, I agree! But I think it will happen. I both fear it and hope for it. So far, the information revolution has consistently performed beyond our imaginations. Twenty years ago people had difficulty grasping that the Internet could be bigger than the telephone system, that everybody in the world would have a cell phone, that there would be cyberwars.
What will make VeeMe happen is mobile, the cloud, big data and social networking.
The mobile revolution means that everything I do is about to be recorded. The cloud revolution means that all that data is going to be accessible together with the data created by others. The big data revolution means that it will be possible to data-mine the information I and others create, finding correlations between the things I do, or between my actions and the actions of others. The social networking revolution means that it will be possible to figure out who influences whom about what.
When all our actions are recorded and analyzed, it seems like a grand challenge, but not an impossible task, to build systems of VeeMes—software that simulates the actions and reactions of people—to mimic the recorded interaction between people.
Let’s say all our interactions of all people get recorded in the cloud. It’s not an unrealistic assumption—just wait until most people on earth have a smartphone, only a few years from now. Now imagine you create a system of AI agents, VeeMes, one for each person. Assume you manage to train this system of VeeMes so that their interaction mimics the interaction between the corresponding people. Well, then you have just created a model system of both people and society. Think of the power in that!
VeeMe will be built from all the collected data about me: my e-mails, social network data, credit card data, health data, tax data, phone calls, location data, and so on. I will also draw on contextual data that can explain why I do the things I do. If it’s raining cats and dogs I’ll probably want to rent a Zipcar instead of going by bike. So VeeMe should collect weather, news stories and more.
To create VeeMe, I index and analyze all my data, and create a model that can simulate my behaviors. I use a mixture of techniques, such as data mining, language analysis, social network analysis, artificial intelligence and so on. I have stored millions of complex stimuli-response patterns, now I am training VeeMe to replicate them.
In this way VeeMe gradually learns to behave like me. And the learning doesn’t stop once VeeMe is created. I let VeeMe join me in everything I do. So my virtual me will become more like me by the day. He grows and develops with me.
I can ask VeeMe about things I have difficultiy remembering. I can ask for “my own” advice. VeeMe will suggest what I would do in a given situation. VeeMe has a perfect memory (perhaps this can help me to avoid repeating mistakes). It can also help me to better understand important events in my life, to resolve conflicts, and so on. It’s so often we forget the details of what people said or, thinking they said something else. Our memories are colored.
Let’s say everybody gets a VeeMe in the future, just like everybody previously got a computer and now is getting a smartphone. Then it gets really interesting. Imagine VeeMes talking to one another. If my VeeMe and your VeeMe like each other, you and I might want to … Next Page »