Self-driving cars, replicators from Star Trek, or even Rosey from The Jetsons: It is an exciting time to be in the field of artificial intelligence with many ideas defining what it means. It also is a popular career path for product professionals. What does it actually mean outside of all the hype and excitement surrounding this term? Procore Product Lead Reza Shirazi talks about demystifying artificial intelligence and what it means to build AI products.
On the nuts and bolts of artificial intelligence
Science fiction has given us many imaginative definitions of what artificial intelligence could be, from robots that clean our houses, cook our food, and build our products, to scary beings that have become sentient and make us think about how we are treating them and each other.
“AI is basically smart machines. If you really think about it, let’s think back on history. Humans have been wanting to invent smart machines for ages. You can look at Greek mythology. There are examples of the people in those stories, thinking about making things that were smart, that would do things for them. So there’s always been this idea in human history of humans trying to do things better with technology. AI is the latest form of that very sophisticated technology that can make humans more effective at what they do.
So AI basically is using all the computational power that we now have, that’s available to us, as technology that can help humans do things in the world better. Whether it’s automating things or giving us insight into things that we can’t discover just by ourselves, AI is really an accelerator and a catalyst for all the things that we want to do with a sophisticated way of doing it.
The first thing that as a product manager, you have to think about when it comes to AI is not necessarily about the technology, but about the problem that you want to solve. With every product that you work on, with every role that you have, as a product manager, you want to get deep into understanding what the problem is, why the customers are having that problem, what would be valuable to them if you come up with a solution, and can that be done? Is it valuable to them? Is it viable? Is it feasible? All those things that could make it something that would work. So, you really can’t start by thinking about the technology.
The playbook for product managers in AI is the same as a playbook for a product manager that is working with other technologies. That’s the first place to start.”
On the ethical implications of artificial intelligence
We see all the benefits to artificial intelligence, but are there any dangers to this technology? Reza shares the ethical questions any product manager in this sector has thought about as they develop a product.
“You’re raising a really critical point that every product manager should think about no matter what product that they’re working on, which is like, is this ethical, is what I’m working on and how I’m working on it an ethical way to do it? That’s this foundational or fundamental question that you have to ask as a product manager. If I’m a product manager working on an AI product, there is the question of, what is this technology going to affect, what are the tradeoffs of doing this AI feature, and how will it affect the human that is part of it? So if I think about the monitoring engineer, in that case, I’m making their work easier, I’m not necessarily creating an additional problem there, they’re able to solve and remediate an IT issue quicker than before.
Now, let’s think about the construction example. You could have some ethical choices or tradeoffs to make. As you’re taking pictures on a construction site, what privacy issues could you encounter as you do that because you want to process those images to solve a particular problem around safety? But you could be impinging on someone’s privacy. So as a product manager, I need to be thinking about, how am I doing that? Or if I have to build out a particular model to determine whether a site is safe or not, how do I ethically source data that would help me build that particular model to where it is actually effective in the field?
So those are sort of key questions to ask as a product manager. It’s a question that we should ask no matter what product that we work on, am I doing this in an ethical way, what are the tradeoffs for the folks that you know, the customers that we’re building it for, and will it be valuable and not cross some ethical line?”
On building relationships between data scientists and the development team
The collaboration of teams is an important part of a product manager’s job. This fostering of relationships is valuable between data scientists and the product development team, especially when it comes to artificial intelligence. Reza shares some ways he thinks about building these relationships and the approaches he takes.
“Working on an AI product, although they’re many similarities to working on any other product, there are some nuances that are different. Working with data scientists and data engineers … is different than working with software engineers. It’s in the name itself, as a data scientist versus a software engineer. Software engineering is typically a little bit more rules-based, and so there’s more pretty predictability in what it is that you’re going to build. When it comes to data science, it is a little bit more scientific. You have these very skilled people trying to replicate the physical world with this model of the world, and in the meantime trying to figure out whether that model will actually work. It’s a little more difficult to predict how long that will take; data science is just more exploratory than rules-based.
Working with data scientists requires maybe a slightly different rhythm in how you are, a slightly different approach in how you work with them. You just have to keep that in mind. It’s the difference between rules and in science; there are just some ways that you need to collaborate with them that are probably different than engineers. You can’t be as rigid about it like, ‘Hey, we got to get this done within a sprint.’ Well, their exploration might take more than a sprint, and maybe you shouldn’t have a sprint for them. But you have certain milestones that say, by this point, we need to have something to share of your research so that we can then make decisions about what we do next.
There’s this greater risk of, will you build a model? Will you build it in the time that you have? And will that model actually be useful? This leads to the second thing that is, when you’re building that model, data plays a critical role. The quality of the data plays a critical role in the lifecycle of building the model. If the data scientists have a good set of data, then they’re going to have the ability to build a really good model. Then you have to think about, do I have the right set of data? Is it ethical in how I’ve represented it? Will it give us the types of answers that we want? Will it actually represent the real world? You could build a model and theoretically, it could work with your artificial data set, but when you put it out in the real world, it might just blow up and not actually give you the answers that you want. There’s a bunch of considerations to think about the data and working with data scientists as you work as an AI product manager that you need to keep in mind.”