Caleb Moses on the Bleeding Edge
Caleb Moses is a Māori data scientist from the Hokianga region whose interests are machine learning, language and automation. Caleb attended the March 2019 Indigenous Protocol and Artificial Intelligence workshops in Hawai’i. Here he talks about his interest in using AI as a tool to create things.
Caleb: My name is Caleb Moses. I’m a data scientist from New Zealand, based in Auckland. I’m working for Dragonfly Data Science, and we are working with Te Hiku Media on an exciting body language technology project. So we built the first speech-to-text algorithm in Teo Reo Maori. So where you can speak Te Reo, the Maori language, to your computer and then it will be able to transcribe what you’re saying in real time. My relationship to AI is that I like to build them.
At university, I did mathematics and when I graduated, I was looking, you know, what are the interesting maths jobs that I could go and apply for. That’s how I learned about data science, how I learned about machine learning. I spent about a year, well, no. I spent about a year studying on my own, pretty hardcore, and then another year trying to apply it in my work and, eventually, found myself working at Dragonfly with Te Hiku.
So, personally, I’m more interested in using AI as a tool to create things. Basically, what you do if you are kind of interested in AI and stuff is you find the people who are on the bleeding edge, and you follow them. You follow them on Twitter. You see what work they’re doing. You see the stuff that’s coming out of the big labs, DeepMind and Google brain and Uber and all of the stuff that they’re doing, Facebook. Then you try to figure out how you can take those technologies, and then use them on your scale because one of the big problems is not just access to know-how.
Because, generally speaking, at least for someone like me who has a university degree and that sort of thing, there’s a lot of resources available online where you can go and learn how to put these things together yourself. I know that for indigenous communities where university degrees are in short supply, that’s not necessarily what could be considered easy access. But at least for me I’ve been able to kind of find stuff online, find blog posts, read them, figure out how to put them together, how to run the models myself. I’ve been able to do that.
But one of the really big gaps between us and Facebook is just computational power. They have so many more computers than we could ... We can scarcely imagine how many computers they have. I remember finding out a few years ago that Google ...that they built this new kind of hardware to do AI models real fast, and they basically ... I managed to find the source that said that their models ... like they have so much computing power that they can run, like, object recognition across all of Google Maps, like all of the street view for the entire world, they can do it in about two days, which is like, yeah. Yeah, there’s no way that a person could do that. There’s no way that a university could do that. Yeah, it’s totally insane. I’ve been excited being here at the conference getting to talk with people who have access to more indigenous data than what I’ve been able to find so far. Te Hiku themselves have probably, so far as I know, the best collection of at least Maori audio, but probably also Maori text now that we’ve assembled language corpus, and I’m definitely looking forward to doing some interesting work with that.
Just a few weeks ago, I was working on a model that basically generated Maori language text. You just feed it all corpus and then it learns how to make new stuff. Doing pretty well, but I think it could do a lot better and, yeah, just more work. More work needs to happen in this area. I think that that’s another thing that, at least as indigenous people, we could really kind of leverage that knowledge that we have about where we come from and to create new things. And not just new things, but new things that only we can make, or at least that only we should make. So, that’s what I’m excited about.
“...what you do if you are...interested in AI...is you find the people who are on the bleeding edge, and you follow them.”
My dream, and I say dream sort of on purpose, I want to see an Indigenous AI research lab that creates things that are Indigenous, yes, but also things that are on the bleeding edge with everyone else. That’s what I want to see, so I want to see us making our own image recognition algorithms, and our own AIs that play chess better than humans and all of that sort of stuff. But also using that Knowledge to kind of create new things, new ways of interacting with our culture, like building new tech, the stuff that Te Hiku are doing now, voice recognition in Te Reo Maori. We could create our own virtual assistance. We could bake them into video games. People could play video games where they have to say and spell in Te Reo Maori in order for it to work. My idea it would be us kind of creating new technologies that are just out there with the best of them. That’s what I think. That’s what I think we can do.
Caleb Moses is a Māori mathematician from the Hokianga region whose interests are machine learning, language and automation.
Before joining Dragonfly in 2018 he worked at Stats NZ in a statistical analyst role focused on privacy, data visualisation and automating routine tasks.
He has a bachelor’s degree and a postgraduate diploma in pure mathematics from the University of Auckland. During his studies Caleb researched mathematical physics and fractal geometry, which contributed to his understanding of the latest statistical models. Caleb is passionate about language. He speaks Japanese fluently and is currently learning Te Reo Māori and Korean.
The Indigenous Protocols and Artificial Intelligence (IP-AI) workshops are founded by Old Ways, New, and the Initiative for Indigenous Futures. This work is funded by the Canadian Institute for Advanced Research (CIFAR), Old Ways, New, the Social Sciences and Humanities Research Council (SSHRC) and the Concordia University Research Chair in Computational Media and the Indigenous Future Imaginary.