Recently I read about an article by Dimitri Bertsimas, who is the Vice Provost of MIT Opening Learning, that describes a vision of what he calls Universal AI:
“Universal AI is a platform that involves a horizontal collection of modules covering the fundamentals of AI with an application-oriented process. “
He further details what are in the horizational modules, and what vertical structures there may be:
“These horizontal modules cover four areas:
- predictive AI using structured data;
- predictive AI using unstructured data;
- multimodal AI combining multiple structured and unstructured modes; and
- prescriptive AI where we try to make decisions.
To complement the horizontal modules on the fundamentals of AI, the platform will include vertical modules of AI + X highlighting the various applications of AI in specific fields. In health care, for example, we could have vertical modules for AI + oncology, AI + cardiology, and AI + hospital operations – to name a few.“
This is such a bold and yet compelling vision. Dimitri is a world-renowned expert in operations research and AI, and his idea of this “T”-type of structure makes a lot of sense. Horizontally you need breadth of knowledge: the fundamental of predictive AI as well as prescriptive AI. But you also need to apply AI knowledge to create real impact in specific areas, hence the verticals.
I am also reminded of the book “The Age of AI: And Our Human Future”, a by Daniel Peter Huttenlocher, the Dean of MIT Schwarzman College of Computing.
Dan also made a compelling case that AI is fundamentally different from previous generations of technologies, in that AI is changing how human view ourselves and our society, and how we perceive the world around us. AI is probably the first generation of technologies that rivals some of the things that we see as unique human.
I conclude with what Dan said in the short interview:
“The AI future is going to happen, but where it will end up is up to us.”