Tanya Berger-Wolf discusses how WildMe uses H2O.ai to support its mission of creating artificial intelligence solutions for wildlife conservation.

 

Speakers:

Dr. Tanya Berger-Wolf, Director, Translational Data Analytics Institute at OSU || Co-founder & Director, Wild Me

Read Full Transcript

Tanya

So I'm Tanya Berger-Wolf, and I am, among many other things, a co-founder and director of the nonprofit Wild Me that creates artificial intelligence solutions for wildlife conservation. The organization started out with a research project. Actually, what started out, it has a couple of origins stories. One of them is a passion project for a whale shark loving individual engineer who decided to create a website to connect whale shark lovers and their pictures.

And that started the original Wild Me and the whale book for Whale Sharks in early 2011, my graduate student at the time created an algorithm for identifying individual zebras from photographs. Part of a bet that I made with biologists that we could identify individuals zebra from photographs with two clicks and the resulting paper captured the world's imagination with many public media posts.

And so that came to attention of another computer vision researcher, Chuck Steward, who really looked at our algorithm at the time and said, we can do better. And he did. And continued graduate students created an algorithm for identifying individual animals from photographs of many different species published in an obscure conference. But despite that, we within two months, we had requests from more than 70 seven zero different organizations.

Can you identify my species from little snails in Hawaii to painted African painted dogs in Botswana? Everybody, it turns out, needed that ability to identify individual animals from photographs, to track them, come them to their social network ranges estimates. And so we really realized at the time that it could be a super useful tool, but it needed to go from an obscure graduate student written code to production level platform that people without any background in technology could actually use for their purposes.

And that's we found wild Me Wild book for whale sharks that had the data technology that we needed and merited with a computer vision algorithms that we were creating to really bring together this new platform wild book.

 

Question

The need was clearly there. I think did a great job painting the the need with the dozens of requests and walking us through sort of the best in the world. I would love to hear a little bit about what makes A.I. a tool that you think can solve this problem if you take us back to to that moment.

You've talked quite a bit about the role photos play from lunch and all the way. I'll let you say it yourself, but we'd love to hear more about what makes A.I. a tool that helps us solve this need.

 

Tanya

The need for data and biodiversity is incredibly pressing right now. As you know, we're saying that we're in the middle of the sixth extinction. And if we didn't, indeed in the middle of a species extinction of that scale, then we need to be able to take action at a global scale and much faster with the current data we cannot do that.

So where do we find sources of data and how do we bring them in, there’re opportunities, the different opportunities? Maybe we can extract more information out of existing data. There are tons of data that's kind of fallow data that's already being used. But maybe we can go back and get more out of it. But there's also the source of data, the images that it's clear that humans cannot just process it and the rate and detail that is necessary to get biological information.

So the time is right. The technology is there. We can we take these massive collections of images coming from a variety of sources and then use the rapidly developing machine learning and AI technology to find these the images that contain animals, find where the animals are in those images and identify species and down to individual animals. So that's where machine learning, computer vision and the AI can really come together technologically to make those data useful.

The other aspect I think of AI that we really underestimate is the ability to make those data personal and to connect people with data, with animals and with each other. When we build the AI technology in Wild Me, we're not only worried about the accuracy of the algorithms for species identification or for individual ID, we worry about the technology being able to connect, to be a good partner for science conservation, to be a good partner for engaging people, to be a good partner to enable data sharing.

We don't talk about the accuracy of the algorithm. We talk about the trade off of human effort with AI kind of lift on the existing baselines.

 

Question

You mentioned the time is right, and I can just double click there. Mm hmm. Right. What does that mean to you as these as these opportunities come to this intersection?

 

Tanya

I think there is a confluence of the real recognition of the scope and challenge of the biodiversity loss, the incredible pressure to do something about it. Right now and the availability of data and the rapidly developing technology. So the time is right, not only because we have the tools, but the time is right, because this is the time to use those tools to address the biodiversity loss challenge.

 

Question

And you mentioned the availability of data. What does that mean? Where is this data coming from?

 

Tanya

Well, the sorts of data that we care about, which is images or it is we decided to do something about, which is images that data are available because we now have everybody taking pictures. Right everybody today has an incredible camera in the in their hands that is available at any moment. And so those are phones right. Something that wasn't there even 15 years ago.

But not only that, we also have all the data that has been unprocessed from the scientific and conservation projects, from the camera traps, which are the motion activated trail cameras, there’re millions and millions and millions of images from that. There's also the the platforms increasingly platforms that allow people not only to take a picture, but to upload it to somewhere where it can be identified and connecting them to other people who are maybe seeing the same species.

So there's all of those data not only people are taking those pictures. There's technology and platforms and ways and pathways for them to kind of make those data available where there is just normal social media or the citizen science platforms or for the scientific projects to bring their images of camera traps or drones and autonomous vehicles in different terrains and in the water on the ground or in the air to really make the data actionable.

 

Question

I would love to talk a little bit about this this concept called democratization. We do have a lot of age to about democratizing A.I. and I wonder just what that means to you and to your organization.

 

Tanya

Well, democratizing AI today is the concept that many organizations are talking about, and I think it means something different, slightly different to everybody. So for me, democratizing A.I. means, first and foremost, making sure that the AI solutions are created with those for whom they are intended, not just for them, but those who are going to use them are at the table throughout the process of creation of those solutions.

The other aspect of democratizing AI is to make sure that AI is created by a diverse set of people. That it's not only one perspective, one viewpoint. One way of creating AI solutions that's represented. And the third aspect of it is that they created solutions. AI solutions are usable and actionable by people that may do not require extensive technical training before they can can apply those solutions.

With all of that, it's not only people. There's a large disparity of where AI solutions are currently created in the world versus where the need for AI solutions is increasingly pressing. And so I think there's the aspect of democratization that's not only at individual level for the people at the table, for the people as part of the process, for the people using it.

But where, where are the nations, where are the places where’re the continents that are represented in this process.

 

Question

So there's a degree of responsible AI as much as democratization of it.

 

Tanya

I think those two are inseparable. I think that I think democratizing AI is part of responsible and ethical AI. And in the process of democratizing AI, we have to make sure that we're doing it responsibly.

 

Question

Another that's sort of a buzzword that a lot of companies use AI for good. Mm hmm. What does that mean? What does that mean to you? I mean, it started as a bet. I guess, but it was also a lot of driving force behind your initiative. So if you would, a little bit about the AI for good and what that means to you.

 

Tanya

I think we all hope that when we're creating a tool, when we're creating a method, that we're creating it for good. Right. So every one of us, of the makers. And so at least it's my hope that this is what everybody has in mind, at least even when they're creating something out of just sheer curiosity, which is how it all started.

I think the emphasis is not only that it's for good in this very abstract way, that it is for social and global good, addressing a social and global problem, which where AI can actually make a difference, in doing that and creating solutions that address or attempt to solve social global challenges requires also extra care in how we do that.

For example, that in the process of saving endangered species, we don't wipe them out by making data of endangered species widely available for nefarious purposes. Making sure that when we provide answers in population sort of growing or shrinking as the basis for policy action, that our numbers are reliable. It also requires that in the process of creating AI solutions, the resource use by these AI technologies does not, in and of itself contribute to the biodiversity loss and climate change.

 

Question

Two, maybe three more questions. One, if you could quantify and maybe also qualify a little bit how A.I. has made a difference in your plan. And I'll ask a slightly different way as well. What if you didn't have A.I.?

 

Tanya

Sorry.

 

Question

So what role has A.I. played in in the impact? What is the difference? It has played.

 

Tanya

Within the world of biodiversity monitoring without AI. It's not so long ago. A.I. Solutions. Machine learning solutions are only now starting to be used in the in the area of biodiversity monitoring. That world is full of numbers with high uncertainty, with data that is unavailable or very, very infrequently available, and policies that are based on data like that quite often and the maybe being the wrong policies or the also hard to monitor, hard to take, hard to assess whether they're actually working or not if you don't have data frequently enough.

And so in in that world where we that moves slowly with high uncertainty and very hard to assess the effect of different actions that are taken. You know, it is increasingly, increasingly challenging to deal with the accelerated pace of biodiversity loss. So to bring AI into this allows us to accelerate both the response, to make the metrics a little bit more precise, to really see all policies working.

Are they working the way that we intended them? And particularly in a world where resources are constrained as they are constrained and conservation, they're really constrained in conservation, the allocation of resources needs. Very good foundation for prioritization, AI can actually do that, can support much more precise prioritization for these kinds of resource, very precious resource allocation.

 

Question

And if we try to distill that back, there's going to be hard lives that you can think of in it. We distill that down to a sentence. It helps us and you can say the full sentence would be great. And then film A.I. helps us help authorities blink. It helps us affect legislators ability to blink. Hmm.

 

Tanya

AI helps us to help organizations put their resources where the most needed for wildlife conservation.

 

Question

Their future plans. I would love to hear a little bit more about the role I can play in your future plans and how H2O might play in some resources or products or things that you're learning about.

 

Tanya

Stuff we're still talking just about wild me, right, because I have big plans for now. I It's so hard for you to talk just about me because I'm a researcher at part. And so I didn't want to talk about the genomics and the whole new field of science, but that's not, you know, where we're going. We can cut all of this out, right?

Yeah. Okay. It is clear that an organization like Wild Me is creating something that is useful for many, many other organizations and research and conservation projects out there. Today almost every conservation organization has it technology, a group that's taking the technology, existing technology, applying it or developing new ones. There's also the disgrace that I'm trying to be more succinct because I know I'm talking.

We have increasing need for technology in wildlife conservation and biodiversity monitoring. Every organization out there, small or large, feels the pressure and drives the pressure. There's an increasing development of that technology. Wild Me fulfills a particular fear of a particular niche in this, we can identify individual animals. We can essentially create that. We're creating the technology that not only allows people I'm sorry, my.

 

Question

Greatest.

 

Tanya

Hope is now. Yeah, yeah. I was like, Yeah, I know, I know. I'm for some reason and I'm not doing a good and good way of selling the future in which I care. With UN and other organizations claiming that biodiversity has a data gap, we have right now a challenge to fill that data cap for biodiversity monitoring. The need outpaces our ability to to to develop the necessary technology.

My greatest hope is that with a challenge this big, there's room for everybody to engage. So everybody who can develop the right technological approaches can can think of them, can come to the table with these organizations that actually deploy the solutions in the field to create the necessary technology. Because AI right now is the answer and H2O the beauty of this AI is that it puts AI in the hands of people who do not maybe have the technological background to create solutions with AI to the problems that they see.

This is what we need right now for the world's biodiversity monitoring community.

 

Question

Last question. I live in Chicago, Illinois, Saint Louis, Missouri, somewhere in middle America. That is literally thousands of miles away from catching the whale whale sharks.

 

Tanya

Mm hmm.

 

Question

Maybe. I've never seen it, Orca. Why should I care?

 

Tanya

The global biodiversity and the loss of global biodiversity affects us all. And it affects one species, probably more so than any other. The humans. We may be the last to go, but the process is already there. So now we can do something about it and you may not care or have even seen a whale in your life, but you've seen other species.

Tanya

You've gone on hikes, you've gone to this. You've I'm sure sometimes reacted some way to nature. Imagine if you could not go on a hike and hear the birds or the rustling of the leaves or see the badger crossing. It doesn't matter which species we all connect. We're all part of the global world's biodiversity and very, very local.

One. We can start by taking a picture or creating the technology, or helping to deploy or helping monitor the species or helping implement the policies, or by just taking these nature works and connecting with nature and remembering that we're part of it.

 

Question

Anything you'd like to add? Anything we didn't cover, anything that.

 

Tanya

Yeah, sorry. A couple of things. So, one at Wild Me we're open source, open collaboration, protected data. We've always been open source and we're always taken care to protect our data of endangered species. But having our code open source doesn't mean that everybody could could contribute. Currently, we're actually building an open source community, the proper open source project, so everybody with ability to contribute will be able to contribute not only by taking a picture, but by creating yet another way of identifying species that we couldn't identify until now or improving the current technology.

So community like H2O that AI of makers of this technology can really participate and contribute.

For more information, you can find us on our website. Wild. Be that org. On Twitter at wild me org.

 

Question

Thank you for doing this. Thank you.