สำเนาบทสนทนา 🔗
Planet Labs SHIFT_mixdown
===
[00:00:00] My name is Robbie Schingler. I'm the co-founder and chief strategy officer of Planet Labs. Planet Labs is an Earth observation company. We are able to build satellites that continuously monitor the planet, and we've been doing this for over a decade, so we have a daily recorded history of the entire planet.
And now with the power of, of AI, we're able to really allow for us to see correlations across space and time. It's a brand-new capability that's helping to digitize many different industries across agriculture and forestry and finance and insurance, and also civil government and defense and intelligence applications[00:01:00]
Humanity has been putting things into space for about 70 years, and it primarily was done really by, by nations. Around 40 years ago, there started to be commercial space, and, and that was to put extremely large, uh, telecommunication satellites up in geostationary orbit. And what that essentially means is a big cell tower that's up in orbit that, that, that signals ping back and forth.
But the last 15 years or so, uh, a lot of things have changed. One is that launch costs have come down, and the ability for secondary payloads to get access to space has made it possible for new entrants to come and build novel services and solutions using space. And two is when, when you get the cost down, you can then take more risk, and you can try things faster.
And so your iteration timeline, your innovation timeline really accelerates. If, if we think about, like, how this transition happened, it first happened from the university community. So in the [00:02:00] 2000s, a bunch of aerospace engineering for-- A bunch of aerospace engineering departments at universities were d- were teaching engineers about systems engineering in space, where they would design and build a satellite in a semester.
And this was phenomenal for people to really understand the electrical engineering, the mechanical engineering, the systems engineering of building this electromechanical device that can actually work in, in the, the harshness of space. And then after a few years, a bunch of those little cube sats, which is like a, a 10-centimeter cube, was sitting in a makeshift clean room at universities all around the world, and they needed to get access to space.
And so that's when secondary payloads really became a possibility. That started happening in the 2000s, and that's what we saw when we were at NASA. We were building satellites. I worked on an astrophysics mission to try to find planets around other stars. My co-founder worked on a lunar mission. And, um, you can do so much so quickly if you give yourself the permission [00:03:00] to take that level of risk.
But when you have really good science, you wanna make sure it works, so you throw more people at it, more time, more money, and it becomes from a $20 million mission to a half billion dollar mission. And something that should take two years, it takes
Load full transcript
eight years. We really saw that convergence of the maker community, this was at the height of like Web 2 when it was fun, and taking that what we know in space to combine those together into Agile Aerospace.
And so we were, we were one of the first organizations to take that leap to utilize that form factor of secondary payloads as the design constraint for our engineers to build a really, really high-performing satellite in a very small package. Over the course of the 2010s, we ended up launching more satellites than any other entity on the planet, and, uh, we designed these satellites to operate extremely low and to make the fleet upgradable over time.
So I think that was a, a bit flip that allowed for people to think of a satellite not as something entirely too precious, but [00:04:00] actually as an electromechanical device that they can engineer themselves, build and iterate and upgrade over time
So I've been in the space community for 30 years. Started in undergrad building technology with friends and small satellites, and fortunate enough to work at NASA as an intern, uh, during undergrad. Um, and I really ended up catching the space bug. It was a, a seminal thing where I ended up finding my, my tribe at a conference at the United Nations in 2001, where, um, I was fortunate enough to be a US delegate to give recommendations on behalf of the youth of the world for the next 50 years of space activities.
And I was one of, um, dozens and dozens of people from around the world, um, people [00:05:00] who, who saw the role of space for the future of humanity in a very similar way. And from there, that's where I decided to combine my social life with my professional life and geek life and live a life of projects. And so side projects become main projects, and I've been fortunate enough to be at NASA for about nine years between grad schools.
Um, and then about 15 years ago, um, uh, left NASA with two of my very good friends and about six other, uh, friends over at NASA to then start Planet Labs.
The thing that surprised me the most when we got started is I thought the space part would be the, the most difficult thing. And in our industry of taking pictures of the planet, that, that's in remote sensing. And that industry really [00:06:00] hadn't found the internet yet. They were still working on really specialized pieces of software on desktops, and, uh, it, it not only has it not found the internet, but it hadn't found the cloud.
So we had to work to create cloud-native geospatial processing standards and capabilities and interoperability to build out this, this ecosystem of remote sensing that you can query through vast amounts of data. And that has been phenomenal in order to then activate a whole bunch of data that was previously really sitting in cold storage, but to make it available so that you can then query through it.
And that's really only been possible in the last, like, two to three years. And now with this rise of large language models and embeddings and, uh, tricks around getting data shapes down and querying through huge, huge data sources, you can find really interesting correlations that, uh, that can, that, that can provide indications and warnings for things about to happen.
And that, that goes across, again, every sector within [00:07:00] society. So the, the, the, the surprising thing for me was that space was not the hard part. The hard part was really building new standards and an ecosystem, uh, then also allowing for people to digitize their workflow takes quite a bit of time, a lot of trust to build with your customers to know that you can back test it and that it can still work operationally, but then they can do a lot more things on top of having more of a digitized workflow.
When we sat down as a group of friends to do this company, seeing the convergence of the maker community and space and the fact that cloud computing was becoming a utility, is we thought about what mission could we do that really has a lot of benefit for, for society? And we chose to do a remote sensing mission With the [00:08:00] mission to image the whole world every day and to make global change visible, accessible, and actionable.
And today, 15 years later, that's still our mission. I'm very proud that we have a business model that is aligned with our theory of change. And I believe that that makes the most kind of like resilient organization and an organization that, um, that can adapt over time based on what, what it's, what's happening in the external environment is when your business model is aligned with your mission and your promise.
We have over time done a number of things in order to govern our organization well. We went public four years ago as a public benefit corporation, and I think that this is really important for, um, technology companies that have general purpose technology to really take into account, uh, the impact on all of your stakeholders when you're making large product decisions or, uh, governance decisions, and to be extremely mindful about how this technology comes [00:09:00] into the world and into the users so that it can be maximize the positive impact of that.
But then there are so many examples that we have actually done in, in deep collaboration with academic organizations and philanthropic organizations. Uh, one that I wanna highlight that I'm very proud of is a mission that was all funded by philanthropy. We found some really good scientists and, and philanthropists that wanted to measure methane emissions around the world, and we came up with the right type of constellation of actors that, uh, could do this, this audacious mission that no one organization could do by themselves.
And that resulted in Carbon Mapper as a nonprofit. It resulted in a partnership with NASA JPL and Planet, where, uh, we were able to shadow as they built the first instrument, uh, at their office. So then for us to then build the second and the next one and the next one
here at Planet, and then to tie that into the, the scaled space operations that we have.
So it wasn't new ground stations or data pipelines or a satellite [00:10:00] bus, but it was actually all part of the same vertically integrated capability that we had that then allows for there to be a new mission with the primary benefit to be for creating a weather service for essentially carbon and CH4. And it's been a phenomenal program.
Uh, we launched the first satellite last year. The next one's going up, um, the next few are going up over the next 12 to 18 months, and it's a perfect example around what a science as a service can look like, where philanthropists can see a market failing and come in to actually develop a digital public good that then allows for it to benefit quite a bit of not just the science community, but all of society to really understand where our methane emissions are.
[00:11:00] One of the programs that I love is our education and research program. And so this was... What's unique about a data company is that you can create a license that allows for you to sell commercially, but then also a license that has more limited usage so that it's affordable or, or very nearly free for a lot of people.
And so we did that with our education research program. And the innovation that comes out of being able to query through a daily history of the planet for the last eight years is phenomenal, and it goes across a whole bunch of things around earth science. But also people are, are finding things that surprise even them.
So there was one in particular that was done by a, um, a think tank, and they were able to find about 120 missile silos in central China that were marked on, on maps as a, as a wind farm for wind energy. And so, but they like correlated it, they did a bunch of research, they published a paper, [00:12:00] and literally the next day, you know, the strategic command of the US military, they said, "Yes.
See? Told you so." There's something about this level of transparency that allows for people to really understand the changing world in a way that allows for, for people to be aware of the plurality of what's happening around the planet and to collectively sense-make together. And that was something that was really just done by one researcher being curious and using these new modern tools, uh, doing a lot of deep research.
And there are just a number of those examples. I think there are over 4,000 peer-reviewed publications that are, that, that have used our data over the last six years. So, uh, there's just countless amount of, of curiosity that's happening in there when you can use these, these tools to... and to be curious around, uh, understanding the changing planet[00:13:00]
Really, we're at this moment right now with large language models and, and, um, and AI tools and technology that is impacting every sector of society. And remote sensing and space is, is just like any other sector. And I think up until now, the big foundation labs, the big large language models that are out there, they're largely human-centric, primarily trained on text written by humans and on, uh, what we write on- online, and it, and it's a bit of a...
It is one lens to look at reality, but it's missing the, the larger living world. And I think that that is where, um, remote [00:14:00] sensing data can be really quite a profound ingredient to allow for these tools to become more aware of, of the physical planet, what's happening on the world every day, and to gain additional perspective to tri- to triangulate truth about what's happening on the planet.
So I'm very excited for these rise of what are called large Earth models, and to pair that with large language models so that w- our tools, when we begin to digitize a, you know, a lot of our operations across society, is aware of ecosystems, of, uh, planetary physics, of where the assets are, where the risks are, so that, uh, we can then really make better informed decisions.
So that's, that's one axis. And then the other axis is, is what's happening in space. You can largely think of a satellite as being a robot that has specialty instruments on it, but they're all gonna be connected to each other So [00:15:00] data can actually pass through one another. There's, there's a lot of compute that's going on board these satellites too, so that you end up having a compute architecture that is highly distributed that, uh, evolves over time.
We've seen this in the computing history over the last 50 years of things that are consolidated on big mainframes to disaggregated on desktops, then back consolidated on these large computing modules. And, and I think that is, that is definitely where we are today, and it's going to continue to, to go back and forth.
And so being able to have these sensors on these satellites, satellites connected to each other, compute that's happening at them, that begins to look a little bit more like, um, if we take the analogy of the body and around how, like, our reality works and how our brain works and how our senses work.
You've got the sensors that are there coming up through the nervous system [00:16:00] all connected to each other that then goes into your, your, your neurons on your brain. But you have a mental model of the world, right? That's your large Earth model, and that's your large language model. And then you feed in all this information, and the things that go against what you were expecting attracts your attention to what's needed the most.
That, uh, I think is that the combination of, of space and AI is, is gonna be really quite profound. That I think allows for us to become way more, as a society, planetarily aware to what's happening outside of our initial field of view. We can become more aware of how healthy and intact bio regions are for agriculture, but then also for reforestation and biodiversity.
Uh, and understanding the connection between those intact bio regions and what that means to weather patterns. We see the impact of quite a bit of our land use policies right now today with extreme weather, and we are terraforming our planet kind of unintentionally in, in most [00:17:00] cases. That I think that there's a bit of an epistemological shift that can happen where we begin to see the Earth as alive, and we can begin to see our society and our culture and humans as part of nature.
And by being part of nature and with our intellect and our collective intelligence with groups of people can become a keystone species so that we can steward our planet in a resilient, in a regenerative manner. So that's my long-term view of, uh, where, where I can see some of these, these affirmative futures coming out of the combination between space and AI technology.
And really what that comes down to though is, is us and our worldview and seeing our place within it so that we can have resilient, regenerative ecosystems for generations to come.