AI-driven digital assistants are fast becoming our way to interfacing with everything. We largely had them in our phones first, then in our homes, but they are coming in our cars, and they are appearing in our offices. Each of these AI virtual assistants is learning with increasing efficiency what we like, what we want, how to anticipate our needs, and, eventually, how to best make us happy. But the big unspoken problem emerging is that each of these things is largely an entirely separate system with virtually no sharing of common information or consistency of experience.
Let’s talk about why that is a problem.
We are complex but consistent
Having separate interfaces for every machine we worked with made sense before we had intelligence in these systems. Having the same controls for a blender as you’d have for a tank would have created problems in both the kitchen and the battlefield (though I can imagine a bender targeting a male demographic that might have sold rather well if it had a cannon and trigger).
But AI virtual assistants are supposed to adapt the system to the human not the other way around and whether you are driving a tank or using a blender, you are the same person. Now that is a pretty wide range of skills, but you could see that an AI virtual assistant that might have trained on your car, which you drive daily, might have advantages interfacing with you if you then drove a tank. For instance, if the AI in the car learned that your go-to curse word was “blast” in the car it could save you from shooting the tank in front of you accidentally if it stopped suddenly startling you. Boy that could be an ugly oops huh?
In effect, what we need is an AI that learns us across all the systems we use because, eventually, it will become the ideal interface into us. In effect it is as much about helping the system or machine better interface with us as it is to help us interface with the system or machine. Much like you wouldn’t arbitrarily change the interface into a car for every driver we shouldn’t be able to arbitrarily change what will become the primary interface into us just because we changed what we were interfacing with.
Of course, the big adverse impact of the current practice of each system or machine having its own AI interface is that each of these must be trained independently of the other. Right now, even if we buy another car from the same manufacturer, we’ll have to train the AI from scratch. That’s nuts, we could have man-years of effort going into that training, you could have a unique relationship with that AI after years of owning the car and all that work will be lost if you decide to buy a new car.
But once trained, and uniquely optimized for us, we’ll want to move that knowledge and experience to everything that can use it. You think about the annoyance of going to a web site you’ve visited before and not having them automatically fill out forms and prepopulate fields based on knowledge that is easily available and of the that they probably already have, this would be tons worse. (This particularly pisses me off with government, medical and loan forms).
The universal AI digital assistant
What is needed is a universal AI digital assistant, one that can learn you cross system and becomes the pervasive interface to us over the years. That would allow it to learn far more about you more quickly, gain intelligence in depth about you more successfully, and raise to the potential for how effective it was in operation significantly. The next Windows or iOS might not be an operating system at all, it could be, and maybe should be, a universal AI digital assistant.
This article is published as part of the IDG Contributor Network. Want to Join?