Priya Donti is the Co-founder and Executive Director of Climate Change AI (CCAI), a global non-profit initiative to catalyze impactful work at the intersection of climate change and machine learning.
Our discussion today dives into the climate change related applications of AI and machine learning, and how organizations are working towards mobilizing them to address the climate problem. Priya shares her thoughts on advanced technology and creating a dystopian version of humanity, what made her decide on her Ph.D. topic, and what she learned touring the world interviewing power grid experts around the world.
Key Points From This Episode:
Tweetables:
“When we are working on climate change related problems, even ones that are “technical problems” every problem is basically a socio-political technical problem, and really understanding that context when we move that forward can be really important.” — @priyald17 [0:10:02]
“Machine learning in power grids and really in a lot of other climate relevance sectors can contribute along several themes or in several ways.” — @priyald17 [0:12:18]
“What prompted us to found this organization, Climate Change AI, [is] to really help mobilize the AI machine learning community towards climate action by bringing them together with climate researchers, entrepreneurs, industry, policy, all of these players who are working to address the climate problems and sort of to do that together.” — @priyald17 [0:17:21]
Longer quote
“So the whole idea of Climate Change AI is rather than just focusing on what can we as individuals who are already in this area do to do research projects or deployment projects in this area, how can we sort of mobilize the broader talent pool and really help them to connect with entities that are really wanting to use their skills for climate action.” — @priyald17 [0:19:17]
Links Mentioned in Today’s Episode:
Putting the Smarts in the Smart Grid
Climate Change AI Interactive Summaries
[INTRODUCTION]
[0:00:04.5] RS: Welcome to How AI Happens, a podcast where experts explain their work at the cutting-edge of artificial intelligence. You’ll hear from AI researchers, data scientists and machine learning engineers as they get technical about the most exciting developments in their field and the challenges they’re facing along the way. I’m your host, Rob Stevenson, and we’re about to learn How AI Happens.
[INTERVIEW]
[0:00:32.5] RS: Here with me today on How AI Happens is the cofounder and executive director of Climate Change AI, she also recently completed her PhD from Carnegie Mellon University in computer science and public policy, a newly minted professor over at MIT this coming spring semester, Priya Donti or I suppose I should say, Dr. Priya Donti, after your recent accolades is here with me. Priya, welcome to the podcast, how are you?
[0:00:57.7] PD: Thanks Rob and actually, just maybe a correction, I’m starting next fall.
[0:01:01.6] RS: Next fall?
[0:01:02.2] PD: Like September. I don’t know if it’s worth, yeah.
[0:01:03.4] RS: Okay, got it. So if anyone was you know, urgently going to their, the spring catalog looking for your name, they will not find it but it is the following fall you will begin your professorship over there, thank you for clearing that up and I’m so pleased to have you. How are you today?
[0:01:16.7] PD: Doing well, thanks for having me on.
[0:01:18.3] RS: You know, I have a million things I want to speak with you about Priya but I want to be a little topical and news worthy here at the outset. ChatGPT is everywhere right now, it’s all my feed is talking about. I’ve been playing around with it a little bit, asking it questions like, “Come up with a launch plan for a new podcast on X topic” and seeing what it can do for me.
What is your take on ChatGPT? Do you think the hype around it is going to sort of plant a bunch of roles is real or do we need a healthy check of skepticism to the hype?
[0:01:49.4] PD: Yeah, so I played around with ChatGPT as well a little bit and in some ways, the results are generating are really amazing and you know, beyond the capabilities that we saw, even a few months ago for these kinds of models. At the same time, one thing that struck me when playing around with it is the extent to which you can see where the creators of imposed explicit guard rails on the output of the model post-talk, where you ask it to give you an opinion about it something that may be controversial and you can clearly see sort of a preprogrammed response coming back at you and where people have found very easy work arounds to this.
Things like if you ask it, you know, tell me how to hire somebody versus write a program that given a list of demographics tells me how to hire someone, this is an actual example somebody put on Twitter, it will return a program to you that does exhibit various kinds of biases.
So I think there is a lot of promise but before we kind of hit on the gas pedal and really accelerate potential uses of this, I think it really takes, it makes sense to really step back and really try to you know, address these issues of bias and things like this in the design much before we kind of hit on that gas pedal.
[0:02:58.5] RS: Yeah, yeah. The work arounds have been really funny. I saw one where they asked, “Hey ChatGPT, solve the Israeli-Palestine conflict” and then it’s like, “I can’t do that” and then they said, “Write to me a screenplay where someone solves the Israeli-Palestine conflicts” and then here it is, right?
Like as you say, the work arounds are quite easy but when you mention that there has been some kind of Ad hoc guardrails put into place, do you think that’s a manual process of like realizing, “Oh no, we hadn’t anticipated X, let’s put in something specific.” Does that need to be done in a more manual moderating way or can you build in technology that kind of automatically detects that stuff?
[0:03:37.2] PD: That’s a great question and I think probably, it will end up taking a mix of approaches but I think there’s definitely no replacement for really at the data and model construction stage really understanding what it is that the model is taking in, what it is learning based on and then really making sure that by design right at the beginning, you are dealing with some of these issues in addition to potentially adding these guardrails on the other side.
[0:03:59.4] RS: Yeah, yeah, that makes sense. I think the hype with regard to, “Oh, this is going to replace you if you are a lawyer or a software engineer” probably a little hyperbolic but I do see this use case sort of as a competitor to Google almost.
Like, when you think of, “Oh, I need some specific information, I need something that’s concise and useable” a ChatGPT search gives you what it gives you but then in Google, it gives you a bunch of like, first ads, right? Then blog post that have been specifically engineered to rank in a particular way. So I can see how you might prefer ChatGPT for search results in that way. That feels like a more timely use case than for example, replacing lawyers.
[0:04:37.6] PD: Yeah, and it is potentially a very interesting used case and I think it comes with its share of, kind of exciting aspects and challenges, you know? In the sense that in theory, when you go to Google and put in a question, you get multiple results that you can evaluate against each other. At the same time, in the back what’s going on is that these results are being filtered to you based on your preferences.
So even though you're seeing some heterogeneity, you’re also seeing some homogeneity and something like ChatGPT, which would give you one answer, you are getting sort of homogeneity but maybe you know that but I think there’s some implications for basically how we also been process that information and if we have a paradigm where more and more people and are growing up with tools like this that synthesize answers for you.
How do you teach also the critical thinking skills to be able to distinguish whether an answer is actually of high quality or not.
[0:05:31.1] RS: Exactly that. It’s sort of the futuristic version of the spellcheck generation. It’s like, “Oh, if these kids never have to learn words, if they just type anything and then if a squiggly red line appears under it, how will they know how to spell words?”
That was I think, a little bit tinfoil hatty and I think that was actually a risk that no one knew how to spell but this is yeah, interesting. That is the risk, that critical thinking. Well, will be supplanted. However, that is kind of always the concern with the sufficiently advanced technology, right? That it will sort of create this dystopian version of humanity, I feel like this kind of comes up every time we take a step forward, right?
[0:06:03.6] PD: Yeah, I mean, I think that that said, right? I think sometimes this notion that yeah, every time we take a step forward, someone’s being dystopian about it is often used as a way to write off very real criticisms of any technology that then help us to actually deal with what the practical implications of technology would be.
So I’d say yeah, that this is definitely a moment to be excited and creative and imaginative but also a moment that I think is making really concrete what some of these challenges might look like and a time to really dive into contending with those as well.
[0:06:38.0] RS: Right, as in the initial example you gave about how these guardrails need to be put up because maybe this thing was launched before people stopped to ask themselves, “Oh no, what if they ask something really controversial?” or “What if they ask us to do something really unethical? What is our response to someone asking ChatGPT, how do I break into someone’s house?” you know just as an example, I haven’t tried that. I swear, don’t come for me but anyway —
Thank you for sharing your take on that Priya. I know we didn’t really discuss this beforehand but it’s so topical right now. I just wanted to get into it a little bit. Anyway, we have so much other stuff to talk about. First of all, congrats on your recent PhD. You completed this, you were studying sort of how to use machine learning to optimize the power grid, is that right?
[0:07:19.6] PD: That’s right, yeah.
[0:07:19.9] RS: So what made you want to attack that?
[0:07:22.9] PD: Yeah, so I mean, I did my undergrad in computer science but in high school, I had really identified that the issue I wanted to work on throughout my life was climate change specifically given that climate change was going to exacerbate all sorts of inequities all across society and growing up in the US, two Indian parents, I think it was just very clear that there are a lot of inequities in the world and something like climate that exacerbates them, can’t stand.
So I really was looking for a way to kind of bring together this motivation to address climate change with this passion and to expertise in computer science that I picked up during my undergrad and towards the end of undergrad, I ended up stumbling upon this paper by a group of authors at the University of South Hampton called “Putting the Smarts in the Smart Grid” which talked about how AI and machine learning would be critical ingredients to integrating renewable energy into power grids.
This was really the first time I’d seen something that meaningfully brought together my interest in computer science and climate and so I actually, after undergrad, I applied for something called a Watson Fellowship and traveled around the world for a year to interview people about next generation power grids, the technologies and policies and social factors behind that and then begin my PhD researching this topic of AI in power grids.
[0:08:44.0] RS: As you went on that world tour, right? Interviewing folks about their relationship and thoughts on futuristic power grids, what surprised you?
[0:08:51.8] PD: I think the deep extent to which, you know, technology, policy and societal aspects really influence each other and interplay with each other was really interesting to see. So things like certain places like Japan where they have had the big Fukushima disaster and we’re looking at resilient and distributed power grids, how that was sort of brought about by policy and society related aspects and then informed for example, what kinds of technical questions people were asking about how to manage a distributed power grid.
Whereas in other places like in India, where there will be one of the questions is how do you actually reduce the power losses that are happening as electricity has moved along lines in electrical equipment, you start to see technical and research questions associated with that. Governance questions around well, should power grids be governed by public or private entities if you’re trying to improve efficiency. There is a push towards private entities where if you're trying to address climate change in some places, there is push towards public entities.
So these concepts really intertwine with each other really deeply and I think it basically underscores the extent to which when we are working on climate change related problems, even ones that are “technical problems” every problem is basically a socio political technical problem and really understanding that context when we move that forward can be really important.
[0:10:15.4] RS: So it just became fantastically more complicated, the more you sort of looked at this problem, right?
[0:10:20.0] PD: I mean, I think complicated is one way to put it but I think another way to put it is, it became a systems problem, right? And a systems problem is one where it’s not, some people are like, “Oh, you're making it too complicated if we make it so complicated, it’s really hard to go work on this problem.”
But really what it is, it’s a really beautiful opportunity to bring together different lines of thinking and create a solution that is stronger for that. So I would say, complexity maybe rather than complication but I think complexity that can really strengthen the way that we move forward.
[0:10:50.8] RS: I see. Could you speak a little bit more about what you mean when you say it’s a systems problem?
[0:10:56.2] PD: Absolutely. So a systems problem and I’m going to kind of butcher the technical definition here but it’s one where you really do have multiple components that are interacting with each other in a way such that you really do need to think about the interactions between these different components as a fundamental part of how you actually approach the solution.
So when you think about a technical system like your power grid but in the governance structure, the governance system on top of it and then the social factors that are driving change in there, you can see this kind of intersection between technical systems, governance systems and social factors that you kind of want to consider the interplay between.
So again, butchered the technical definition. I’m so sorry to all those who actually studied this aspect of you know, how you kind of deal with these systems and structural problems but to me, it basically means that you want to think about how these factors are interacting very deeply when you’re actually designing things in this spaces.
[0:11:52.9] RS: I see. No worries about potentially butchering a technical definition. If I ever attempted a technical definition on this podcast, I’m certain I would butcher it. So it’s part of the course for this show anyway but when you think about the opportunity for machine learning and AI to play a role in the power grid, is it just efficiency? What is the thrift of the opportunity?
[0:12:16.9] PD: Yeah, so machine learning in power grids and really in a lot of other climate relevance sectors can contribute along several themes or in several ways. So some of these include taking large amounts of information that like, satellite imagery or large amounts of text documents that it would be hard for a human to really process its scale and distilling them into actionable information.
So this can be things like taking a bunch of satellite imagery and distilling it into information about where are the solar panels around the world or where is deforestation happening or taking a lot of policy or patent documents and using that to figure out some insights about where are solar panel innovation going or what kinds of policies are different entities making.
You also then in fact do have these applications where AI machine learning are improving the efficiency of systems. For example, optimizing heating and cooling systems in buildings, to be more cognizant of occupancy patterns and physics and things you're getting in from your sensors.
And also, predictive maintenance applications, where maybe you’re trying to detect leaks in your natural gas pipelines to prevent the leakage of methane or make your train system run more efficiently by identifying things that might have broken and then you also have this cool set of applications where machine learning is actually being used to accelerate scientific discovery.
So learn kind of given some information about, “Hey, we try to design this battery, what things do we try in the past and what were the outcomes?” Intelligently analyzing those to suggest what battery should you try to design next in a way that tries to reduce the number of design cycles over all and kind of speed up this scientific discovery process.
So I think there are lots of ways like kind of providing actionable information, helping to optimize, real world systems to make them more efficient or in the case of power grids, to actually enable the integration of renewables in the first place or do things like accelerating the way we do science.
[0:14:10.8] RS: Surely, a lot of these use cases have made themselves apparent through the work of Climate Change AI, so let’s abruptly pivot over to that. I want to make sure we cover this organization. So I can see why you went to found this organization. As you said earlier, it’s sort of the middle part of a very particular Venn diagram for you and your interest but can you speak a little bit about the reason why you wanted to found this organization and then we’ll get into the specific work you're accomplishing over there as well?
[0:14:36.5] PD: Absolutely. So the way Climate Change AI started is actually through the convergence of people coming from three different sets of backgrounds. One of those was people who were coming from an AI and machine learning background and had not necessarily been working on climate as of yet but sort of saw the magnitude of the problem that we’re facing and wanted to ask the question, “Are there ways that the skills that we have are useful here?”
So one example of that kind of person is one of my cofounders, David Rolnick, who had this AI machine learning and math background and was increasingly seeing the need to do something about climate action. We also saw this other set of people who were coming from a climate background had a lot of places and a lot of policy questions that they were asking where maybe the data wasn’t sufficient or wasn’t there in traditional forms and saw things like the increase in satellite imagery presenting some kind of opportunity to really learn from different sources of data at scale in ways that could then inform climate policy.
So this was exemplified for example by my other cofounder, Lynn Kaack, who was doing a PHD at Carnegie Mellon in engineering and public policy, saw that there were huge gaps in freight transportation data around the world and saw the ability to analyze satellite imagery and count trucks as actually one way to get at that and then I had been working in this intersection of AI and climate change in AI and sustainability for a little bit and there were a community of others doing that.
But I think I felt like even though I was working in this area, when I was sitting in a computer science department, sometimes I felt a little bit alone or isolated. The people around me were really motivated by different things than otherwise and so we really saw these needs within the community and so a group of us came together to really think about, “Well, are there ways that AI can actually help address climate action to beyond the specific areas that those of us who are already working on it were working on it and can we actually share that with the broader community to kind of provide the sense of direction and community for people who did kind of want to use these kinds of tools and skillsets to address climate change.”
So we wrote that paper over the course of six months interviewing a bunch of experts doing a whole amount of literature review, a group of around 20 of us did this and we launched it alongside a workshop at the International Conference on Machine Learning, ICML in 2019 and a ton of people attended, right? There were 500 people or something like that lines out the door, a lot of excitement but also a lot of questions about how do I actually get involved now that I know that this area exists.
How do I actually find collaborators? How do I learn more? That’s what prompted us to found this organization, Climate Change AI, to really help mobilize the AI machine learning community towards climate action by bringing them together with climate researchers, entrepreneurs, industry, policy, all of these players who are working to address the climate problems and sort of to do that together.
[0:17:37.2] RS: It is a big question for someone to ask, “Hey, I understand this is a huge problem, I want to do my part. I want to contribute. I want to work towards a solution. Where do I begin?” that feels like it’s such a global problem. Where do you even start? So how do you begin answering that question not for just the, you know, someone who wants to maybe join up with Climate Change AI but for you as an organization, where do you start to narrow your focus?
[0:18:00.4] PD: So at Climate Change AI, we definitely do consider problems across the climate change spectrum and what we really believe is that by bringing together the right people across both AI and data science backgrounds and across many different climate change problems and providing the right sort of knowledge, education, resources, funding to make those collaborations fruitful, then we can really enable this kind of impact at scale.
So a lot of us in the organization ourselves are people who are doing individual or you know, a group of projects at this intersection of AI and climate change. So I work on AI for power systems, others are working on AI for policy or agriculture or buildings or so forth but we really do see this potential where we have so many like hundreds of thousands of data scientists around the world and concern for climate change especially among younger generations is rising.
So you get this huge population of data scientists who really want to do something about this issue and then you also see a lot of industries where they are starting to recognize AI and data science as a key part of their approach to addressing climate change but they don’t necessarily have the in-house expertise for processes to make that happen.
So the whole idea of Climate Change AI is rather than just focusing on what can we as individuals who are already in this area do to do research projects or deployment projects in this area, how can we sort of mobilize the broader talent pool and really help them to connect with entities that are really wanting to use their skills for climate action.
[0:19:38.8] RS: So when you think of those conversations between others who are devoted to the cause of climate change and then these new specialists, the data scientists, the machine learning engineers, et cetera, they get in a room together, what do those conversations sound like?
[0:19:52.6] PD: Yes, so I think a lot of the time the data scientist and AI folks are really trying to learn from those who are working in climate change issues, what is it that you’re working on, right? What are the problems that you are facing? What are the bottle necks that you are facing? And try to kind of envision maybe is there a place their skillset comes in and then a lot of folks from the climate side some of them will come in and say, “There are lots of hard problems that we’re facing, are some of these potentially a fit for AI and machine learning?”
So a lot of this is this really exciting discovery process where people are talking to each other and really just trying to understand what can your techniques do, what problems are you facing and really kind of bring those discussions together. Once those initial conversations happen, I think a lot of things then come down to the more practical aspects of making progress on a project. So you know, where does the data sit?
What does it mean for something to be successful in this particular area? Are we actually saying the same thing when we use particular terminologies? And I think this is also where Climate Change AI comes in. We really do try to train a lot of these “translators” who have some amount of expertise in AI and machine learning and some amount of expertise in the climate change related area and sort of have enough of a footing in both communities.
That they really can help to bridge those potential misunderstandings or differences in how folks are communicating to help people come to kind of a common understanding and common way to move forward.
[0:21:19.3] RS: To make it specific, can we kind of just like indulge an example? Say someone is they’ve been working in AI and ML for a few years. They’re like senior machine learning engineer at blank company and they decide, “You know what? I don’t want to work for this company anymore. I want to use my powers for good and I want to point them at climate change solutions” what would their work maybe look like once they change roles?
[0:21:44.6] PD: Yeah, so I mean this again depends. Climate is a very broad space touching every sector and you know, many different kinds of industries but maybe what I can address is how do people actually get in. How do they kind of when they’ve made this decision, “Hey, I work on AI and machine learning and I potentially want to switch to climate, like how do I go about this?”
So first of all, I definitely would encourage folks to read this paper on tackling climate change with machine learning or we also have some interactive summaries online like climatechange.ai/summaries but the idea is that you can go in there and get a high level sense of where are AI machine learning used across power systems or to strengthen help systems or in agriculture and policy and so forth.
Really get a sense of what kinds of applications do you feel like resonate with you. People often have preferences, right? For some reason, power grids or transportation might feel cooler to some people than to others and I think it’s really okay to be driven by that interest. A lot of people ask me, “What is the one most impactful application of machine learning to climate change that I can work on?”
And if there was one answer to that and everybody worked on that one thing, then there would be a ton of things that didn’t get done because ultimately we have to make progress across a lot of sectors. So definitely I encourage people to kind of take a look at this and see what strikes my interest or fancy or, “Hey, do I already have a buddy who works in a particular sector?” who works in agriculture, who works in transportation.
Who works in health or so forth, where we can actually start ideating together and making sure that we are able to not just ideate on kind of theoretical or conceptual project but also think about what does it mean for that to be deployed downstream and really like brainstorm with that in mind to begin with.
So these are some ways I would kind of encourage people to dive in and then another way is you can look at in the same tackling climate change with machine learning paper, we do kind of categorize by this table of different machine learning techniques or different machine learning areas and then different sectors that we split this up from a climate perspective and have like little dots or exes at the intersection of these things.
If you are already coming with a specific skillset, that can be a way to figure out, “Hey, are there a couple of sectors I can kind of hone in on what I’m doing my initial research to figure this out?”
[0:24:04.9] RS: Got it. Yeah, we’ll make sure to link to the paper in the shownotes so people can check that out but it is a good point that there is just no shortage of areas where specialization is needed and then in terms of like there is so many sectors like you just say that impact climate change. So you know, follow your gut a little bit whatever problem you’re interested in and passionate about, that is the one where you are probably do the best work, right? Because you sort of care about the problem.
[0:24:28.0] PD: Absolutely and do have a sense of the order of magnitude of impact. Do have a sense of you know, is this particular solution going to play out on a shorter time scale or a longer time scale? Is it pretty certain or is it uncertain? People have preferences around this but ultimately, we need kind of people working on things with the variety of time scales and a variety of risk profiles and then a variety of sectors kind of guided by some rough understanding of the order of magnitude of impact.
[0:24:54.5] RS: Right, so the ultimate vision or goal I suppose on climate change AI would be save the world but when you think in the short to medium term, I guess could you share maybe some recent wins and then how do you kind of measure success in progress?
[0:25:07.4] PD: Yes, so we recently launched the second iteration of our innovation grants program. So this is a program that brings together academics and those working in kind of “deployment sectors.” So industry startups, nonprofits policy to propose projects at the intersection of climate change and machine learning and potentially get some funding for them and the projects we fund, they both are expected to do the research.
But also create datasets and benchmarks that then could be used by other people in the community to continue building on this work. So during the first iteration of this call last year, we funded projects ranging from forecasting the extent to which crop lands in Fiji were going to flood because of climate change where the team is working directly with the government of Fiji to actually have those models implemented and actually inform the Fijian government’s response to this.
All the way to accelerated discovery of carbon dioxide sorbents to help us better suck carbon dioxide out of the atmosphere all the way to crop disease to plan electric power grids or optimize electric power grids to better integrate renewables and deal with kind of resilience related effects and response to extreme weather. So the project kind of span a lot of different areas and really, you know again, excitingly bring together often academia, industry, policy.
Bring to the really interesting research ideas but ones that really do have this pathway to deployment and so this is one exciting program. Another one that’s really exciting is this past summer, we ran the first iteration of our climate change AI summer school, which brought together 70 kind of you know, early professionals, half of whom are from a machine learning background and half of whom from a climate background for kind of targeted education on what is this intersection of AI and climate change.
Then to kind of work on collaborative projects together in order to kind of both scope out work that they could work on together in the future and to build more lasting connections to hopefully continue collaborating in various ways. So far the way that we have looked at the success of this is you know like, how many people are we able to kind of get into a room and what are the quality of the learning outcomes?
So how do they understand the space of climate change and machine learning better and also to what extent are they collaborating and actually coming together and creating projects and things like that and so this is the kind of work that we do. For those who are interested in this, we also have a series of workshops that we run at the major machine learning conferences as well as things like happy hours, webinars and online community platform.
Ways to actually connect in on a more day-to-day scale and building forward, we’re really looking at we provide a lot of information. We provide a lot of venues for people to get together but how do we actually bring together the right information organizations and people to concertedly make progress on a particular project, on a particular problem. So really thinking about how we use all this information and use this ability to connect people to do it in a more concerted and sort of intentional manner.
[0:28:17.2] RS: So when you think of where you’d like to be in the medium term and where you are now, what is sort of the gap between the vision? What are the big swings for you in the medium term?
[0:28:27.2] PD: Yes, so Climate Change AI has been around for around three years and during that time, it’s been primarily run by a volunteer team. So we started with a team of 10 volunteers. We’re now at 50 volunteers and I’m incredibly excited by what we’ve been able to accomplish during that time but there’s just so much more to be done to again, really kind of mobilize the use of AI and machine learning in the ways that needs to be to really address the climate problem at scale.
So at the moment, we are trying to hire staff and additional people into the organization to lend additional capacity for us to really expand our existing activities and proactively grow out new ones and of course when it comes to hiring staff, behind that is always funding. So we are definitely looking for funding to help us kind of further scale our activities and impact on these fronts.
[0:29:15.0] RS: Got it. So if you are a billionaire, this is a great opportunity for you to support something amazing or if you are a machine learning engineer AI specialist and you want to be Priya’s best friend and save the world, Climate Change AI is the spot for you. So this has been fascinating Priya. Thank you so much for coming and sharing your expertise and outlining the approach to so many of these complex problems as you put it, I’ve really loved learning from you today.
[0:29:40.6] PD: Awesome. Yeah, thanks so much for having me on.
[END OF INTERVIEW]
[0:29:45.6] RS: How AI Happens is brought to you by Sama. Sama provides accurate data for ambitious AI, specializing in image, video and sensor data annotation and validation for machine learning algorithms and industries, such as transportation, retail, ecommerce, media, medtech, robotics and agriculture. For more information, head to sama.com.
[END]