Why Do You Need An Ontology?
Why do you need an Ontology?
The question Alex Karp proposed to Product Manager Eran Witkon recently in a Palantir Youtube Video. Here's how the exchange went:
Eran: “Before joining Palantir, I was the CTO of various organizations. The last one of them was the CTO of the Prime Minister Office in Israel, where I actually met Palantir for the first time. Kind of as a CTO, my role was to kind of identify the right tech that fulfills the business”
Alex Karp: “Just a minute, many people incorrectly think of CTO as a somewhat, you know, erudite profession, which has obviously changed quite a bit in America. But when we met you, when you were doing your work, one of the interesting things was you were very far ahead of the world on it. not being just like stitching together things that may not work. It was very operational. So CTO in your context, it basically meant getting operational products to the front line.”
Eran: “Exactly! And I think part of my mantra as a CTO is when an I.T. professional in your organization is deploying their people, doing something that they know other people outside of their organization can do better. They're not deploying them in the right place. And my view was that we need to use the people, that we have to basically close the gap between the off the shelf product and the specifics of the business need. So it needs to be very, very connected to what the business needs and but also very connected to what I can do and I cannot get outside.”
Alex Karp: “What you see in America now is a move from CTO that's maybe not adjacent to the business to a CTO that defacto is driving business decisions. And you were, you know, Israel, in fact, in many ways was on the forefront of that transformation.”
Eran: “Yeah, I agree with that. Like the fact that, if the CTO is not just, you know, writing papers but actually knows how to talk to the real developers, get them to kind of implement the vision and go and test it with the, with the business. Almost like the forward deployed engineer, you take the I.T. professional and you put them in the business division and they see whether it works or not. That's that's a key part of being successful.”
Alex Karp: “So given how operational you are, why are you in love with a very academic term called ontology?”
Eran: “I think that the real thing is that we've seen a lot of companies investing millions of dollars in time, in kind of building software or buying software for kind of a niche use cases. But the end result of that doesn't accumulate to like to a better business outcome. And I think what's unique about the ontology and Foundry in general is that it's not like an I.T tool and it's not an end user product. It sits in between, it sits in the middle between and connects them.”
Alex Karp: “Given, it's a high end academic hard to define topic that in your view can define whether a business succeeds or fails or thrives. What is an ontology? What is Palantir's Ontology? Why is it important? Why is it differentiated?”
Eran: “To me, it's the intersection of three disciplines in the organization. It's the business. Language is the way the business analysts understand the way the business works, connected to the IT person that knows the data and needs to basically hydrate that these objects or fill them with the real data. So the model doesn't become theoretical. It's actually the real data that the business has. And the third one is the data scientist that doesn't work in the lab.”
Alex Karp: “Most people may believe they have something like that. What is the technical challenge that it's solving that is hard to solve that may not exist without an ontology?”
Eran: So people have invested huge amounts of money in building these data lakes, but they still need the IT professional to run the queries for them. So they don't get what they need out of that kind of effort. And with Palantir, with the ontology, you actually get a much, you know, a usable thing that could self-serve the business that console serves, the business professional and actually work with that. And we've seen that in various places.”
Alex Karp: “Would another way to put this be that many people assume that the read write function or the decisions to data like distinction. Meaning you can read and then write to the underlying database or that you can make a business decision and then change the underlying way in which the data are structured. I.e. controlling how your business runs, something that actually cannot run without an ontology.”
Eran: “Yeah, I think the ontology, the way we think about the ontology is, not just through the nouns, but also through the verbs. The fact that you can actually codify the swapping of an airplane and then not just being able to perform that in your screen, but also having that to write back to the system.”
Alex: “Why is that difficult? So IT professionals will look at this and understand why this is difficult. Non-IT non-technical people may not completely appreciate why this is very difficult to do. Why is this difficult to develop an ontology that gives you a rewrite function across a large organization?”
Eran: “Starting just from the fact that you have an enormous amount of systems that, you know, each one of them has their own view.”
Alex: “Their own view, meaning in simplistic terms, each part of the organization has its own ontology.”
Eran: “Yes, exactly. But the CRM department, the operation department, whatever, the customer service department, each one of them has their own view of that ontology. But they don't speak the same language. And without connecting them into one place, you can actually apply the CEO strategy. So imagine the CEO wants to change the balance between customer service and cost of maintenance. How do they do that? Like today, they will send an email to all of these people hoping that they will apply that each one in their own systems or actually around the system with an ontology, they can actually change the weights that will impact the way the recommendation is given to the end operational in the field.”
Alex: “Maybe a more layperson's version is without an ontology. You never are without an ontology, you just end up with a thousand ontologies which mean your read function is broken. Because when you try to read what is going on with the business, you get disparate, highly inaccurate data sets because they're created in each one's own framework, i.e., an ontology. And the right function doesn't work because they're all disconnected, actually, even if they're connected because even if they're connected, the underlying worldviews are disparate. And therefore you cannot control your own business."
Thanks for reading the article. If you'd like to get in contact, please @ me on twitter here or email me at email@example.com. You can join our Palantir Facebook group here to participate in community discussions, polls, and more. You can check out daily palantir audio content here.