In HBO’s latest hit, Westworld, the creators push the boundaries of what can be accomplished through artificial intelligence. Humanoid robots (or “Hosts”) interact with guests in a decadent theme park where anything goes and rules don’t apply. But while the world of robots escorting us through the Wild West may seem like a dizzy daydream, advances in Artificial Intelligence (AI) technology are becoming increasingly real.
Saman Farid is the head of Comet Labs, the first venture fund dedicated to developing intelligent machines by investing in B2B robotics and AI-based startups. His vision for the future of AI takes us everywhere from improving building construction to revolutionizing medical devices. Propel(x) had the opportunity to speak with Saman about the future of AI in traditional industries and how AI is changing our world.
Propel(x): Tell me about Comet Labs.
Saman: We started out by investing in a lot of AI-related technologies, and we saw so many amazing leaps in terms of what AI has made possible. Computers were getting better at understanding the physical world and making decisions around these interactions. At the same time these interactions are helping us rethink the way we do things. We saw a lot of those changes being applied to fields like advertising, financial services, and sales, but we noticed that the majority of industries in the world weren’t yet getting any of the benefits from advancements in AI. If we look at the GDP in the world, 80% of it is in things like agriculture and construction, healthcare, and a lot of fields that have really been very minimally affected by the advent of the Internet and connectivity, let alone artificial intelligence and machine learning. So we made it our mission to invest in the AI technologies that will impact these untouched industries.
Propel(x): And what was your initial charter?
Saman: The initial approach was “Let’s run an experiment.” That was the charter. And the experiment was–assuming money is not the biggest bottleneck for AI-related technologies–what can we bring to the table to speed up the application of AI to the physical world?
Propel(x): That sounds like a huge challenge. Which areas are you focusing on first?
We saw so many clear needs for the application of AI, and we recognize that there aren’t enough people working on its development. We began to investigate why that was the case. First is the dearth of trained talent. The number of talented people who can build mass machine learning platforms at scale is very, very small. Secondly, the people who are trained in these skills generally are only familiar with a few industries. So we decided that if we really are all-in on AI, then we really need to contribute more than just investment to the problem. And so we started scouring the earth for the people who could build these things and the people who can use these things. Every month we put together events and summits where we really try to understand: what are the things in a particular industry that AI, machine learning and robotics can change?
Propel(x): How do you think robots and AI can impact more traditional industries?
Saman: Let’s take the construction industry as an example. Every phase of the construction industry could benefit from AI–from start to finish. To begin with, AI could really help with the design phase of a construction project, which currently is s extremely manual, slow and expensive. Basically, architects spend time drawing out different designs based on a brief and usually provide two or three options for each project and then the client has to choose from these options. They don’t know what the financial implications are, they don’t know what the construction costs are. Then we get to the actual building process, which is another completely undefined problem. People will build things and then realize they’ve missed something, and have to tear down and rebuild. It’s not surprising that it’s estimated that 30-40% of construction cost is waste.
Through AI, we’ll be better able to observe what’s going on at a construction site. Soon we’ll be increasing the accuracy and the ability of the tools to interact with the physical hardware and physical space, while also integrating the design process into the construction process to close the loop on iteration so that the architects are involved and can quickly give feedback. The building industry is gigantic but hasn’t changed much in the last 150 years. It’s just mind boggling the opportunity that’s there.
Propel(x): You have a very interesting international background and perspective. How has that perspective and Comet Labs’ international presence given you added insight?
Saman: So I spent a lot of my career in China, and most of the Comet Labs team in the US has had experience working in China. We also have a team in Beijing. Being international has been immensely helpful for us. First, AI is definitely a global phenomenon. It’s not like social media, where it may work very well in one area, but it may not translate very well across borders. Things like artificial intelligence are really efficiency driven. And efficiency-driven endeavors are benefited by large, difficult, complex environments. And the two places in the world that have a lot of those complexities are the US and China.
Propel(x): Are you working with universities and AI programs globally to try to identify talent and educate them on the opportunities available in old-line multi-billion dollar industries?
Saman: Absolutely. One of our goals is to expand our reach beyond investment. Right now we have a technology advisory board that consists of professors at universities all across the US and China. We also have a scouts program where we have young people in labs all over the world who are keeping us up to date on relevant academic advances. We definitely don’t see ourselves as experts in any sense of the word. But we think of ourselves as aggregators, and if we can bring all of our people together, we can make a change. We’re currently in talks with one of the largest universities that is focused on automotive innovation, and our goal is to build a self-driving car laboratory with them and a few other corporates in this space. This lab would bring invite startups and students to work on projects and be coached not only on the technology side of things, but also the commercialization side of things, which is something that we have experience in.
Propel(x): You said that just like the mobile revolution, some of these things are actually around the corner, like we’re not talking 20 years, 30 years. What is around the corner? How are most people going to be impacted most immediately by these technological innovations?
Saman: I think it’s hard to predict to be completely honest. But I think that really what we’ll see is gradual changes in terms of the things that AI can do. So, in most cases, that’ll start off in a supplementary capacity. So AI tools will be used to augment humans and give them increased scope and breadth of access and knowledge.
At the same time, there’s the data acquisition side of things. So we’ve invested in a lot of companies that are finding all kinds of ways to gather more and more data and are building sensors and chips to do so. We care about the collection of data because that’s what will help us build tools that assist in decision-making.
So gradually as humans start to do those things in combination with computers, computers will learn those skills from us through things like reinforcement learning and expert systems. And gradually computers start to develop more and more of an intuition about those things. I think that’s the next step of the evolution. Gradually computers are going to be able to adapt to more and more diverse sets of use cases.
Propel(x): Could you give me an example of a company in your portfolio that’s got you very excited and that you think could have a huge impact?
Saman: I’m excited about our entire portfolio but one example we’re really excited about 3Scan. Essentially, they’re a data input system that is slicing human tissue. Previously, the way that you would look at a biopsy is that you would slice it, dye it, wax it, and put it under a microscope. Then a doctor would look at it and subjectively evaluate that sample. This is an inefficient and not very accurate process.
3Scan developed a system where they can very quickly–at about 2000 times the speed–take any sample of human tissue, slice it into very, very thin slices, digitize the data, build up a 3D representation of the scan, and then quantify the scan. So they can say, “This is how many blood cells are in there. This is how many blood vessels are in there. This is the count of this type of number. This is the density or the clustering of different kinds of blood cells.” They quantify the data. But because they can do it at a rate and speed that’s so much cheaper and faster than the current process. And it allows a doctor who could previously only look at two samples per day, to look at as many 2000 samples per day and interpret them accurately because of the improved amount of information they have to go by. Now what 3Scan is doing very quickly scales up as they collect ever growing sets of data. They will be able to provide these analytics and insights to research institutions and to drug developers. Over time they will also be able to make the diagnosis themselves. So essentially, they will be able to take any sample, quantify it, interpret it and give you information about it.
Propel(x): These examples are great. They really bring what’s happening in AI to life. How about an example on the construction side?
Saman: Sure. Another is Shaper Tools, who are building a handheld CNC machine. They use computer vision, which allows their tool to accurately place itself within a work piece and cut things 100% to specification. All you have to do is upload any kind of 3D rendering or drawing and what you want the piece to be made out of, for example, wood. From there the tool knows where it is vis-a-vis the piece wood, as well as what you want it to cut and it will automatically develop a cutting path. Once you turn it on, you just can make a general motion, and even if you’re the worst carpenter in the world, it will fine tune its blade and the place that it’s turning, and will make sure that every cut is 100% accurate every single time. And if you go out of bounds, it’ll just lift the tool and you won’t have any damage to the workpiece.
Propel(x): Tell me a little bit about the other types of support you give your startups.
Saman: We are increasing the way we support our startups everyday. The number one thing we help with is inside knowledge of large industries and corporations. Very technical founders don’t generally get exposure to old-line industries and we act as a go-between and help them understand how to successfully work with these large companies and create products that would work for these companies and are priced appropriately. Through our corporate outreach efforts we’re now connected to about 50 or so corporate partners who are willing and enthusiastic users of these technologies and who want to become customers for our portfolio companies. This way we are able to connect our portfolio companies to their potential customers earlier, and help them shorten that product iteration and sales cycle.
The other thing that we’re doing is putting a large entrepreneur in residence (EIR) program in place, where we’re going to have 10 to 15 EIRs come in and basically give them flexibility and time and space to work on different ideas and then spin them out into companies when they’re ready.
Propel(x): If your efforts are successful, how will the world be better in 20 years?
Saman: If we are successful, people will only have to work a few hours a day to meet the needs of production. Instead, humanity can spend more of its time and energy on what they are uniquely positioned to do. Explore, learn, and grow. People can dedicate time to research, to exploring the oceans and outer space, to understanding medicine and and psychology, to building a civilization that is no longer purely about resource allocation, but can be about building peace and developing the abilities of humanity. It seems far fetched, and AI is definitely not the only contributing factor to make this possible, but my hope is that as machines can take on more of the workload, humans can focus on the more meaningful aspects of existence.