Share data to save the planet, urge Subak founders

Are non-profits the answer to the climate question?

The self-declared mission of the world's first non-profit climate impact accelerator, Subak, is no less than to save the planet with shared data.

Subak selects, funds and scales non-profit organisations that aim to combat climate change through shared data, and provides access to the intelligence, funding and business acumen that those organisations need to grow.

Two of Subak's Cohort Zero, Dan Travers and Jack Kelly of Open Climate Fix, explain why data is now the critical resource that must be shared for the common good, and how they are part of a data cooperative driving insight into humanity's greatest challenge.

"There are thousands of researchers out there who would really love to help fix climate change," explains Kelly. "Mostly machine learning researchers, students at all levels and researchers in commercial organisations. They would relish an opportunity to search for ML problems that, if solved, could have a large, positive climate impact.

"The problem is that they find it really hard to get their hands on the right datasets and they would find it hard to understand exactly what industry needs. Even if they do come up with some amazing research paper, that's no guarantee that industry is actually going to adopt it."

Together, Travers and Kelly founded Open Climate Fix, one of Subak's founding cohort of small non-profit start-ups who share a desire to increase the impact of their efforts to help humankind clean up the mess we've made.

Kelly continues, "We're trying to bridge the gap between those thousands of researchers on one side and industry on the other. Trying to get data to the researchers so they can do their work, and then once the researchers have produced great algorithms, at least build a proof of concept that industry can actually use to evaluate if those great ideas work in practice and reduce emissions."

Travers sets out the importance of not just connecting data, but ensuring its quality and making it searchable. He uses the example of Ember, another of Subak's founding cohort, to illustrate.

"There are subject matter experts in particular fields, such as Ember. They're experts in coal and gas emissions. Ember have been working this data for a long time, so they're connecting up various data sets from network operators [and] various publicly available data sources. But those data sources are potentially not very easy to work with and not connected - so they've done the work to connect those data sets, they publish reports on it, and they push the results from those reports out there."

The problem is that it's very difficult for other researchers and organisations to pick up the baton.

"The data is often left not joined up to the rest of the world, to allow someone else to do more research and gain more knowledge on top of that. What we're trying to do is take data such as Ember has got - maybe you combine that with the solar energy data that we've got in Open Climate Fix - and say, okay, how can we allow a third party to access this data really easily?"

Subak wants to make all of this data accessible, in the manner of a search engine.

"If you look up coal emissions, we don't want you to find just any old coal emissions power set; we ideally want you to find the most useful one, the most connected, the most cleaned, and so forth: the most valuable data set."

"Being non-profit sends a loud signal that we're not competing with you"

Like all the best ideas, it's simple and intuitive. So much so that an obvious question to ask is: why exclude private industry and focus on non-profit organisations? Travers explains:

"Subak is working with non-profits partly for clarity of the vision - all the people in Subak have got the same mission so it's much easier to allow the data sharing. There's also a fair bit of VC available to for-profit companies, so I don't think there's as big a funding gap as non-profits have to deal with. What we want to do is accelerate those companies that are really early stage. We're going to bridge the gap between an idea and something that becomes commercially viable. "

Kelly emphasises the importance of trust when opening up not just data, but ML models.

"Open Climate Fix are developing new forecasting techniques, and there's money to be made from that, but we think the best way to have maximum climate impact is to enable all the other forecasting companies. So, we're upstream of them. We have really talented machine learning people on staff; other forecasting companies don't have many and they don't have the time to do deep research into taking these ideas from natural language processing and applying them to forecasting.

"One of the things we're doing in Open Climate Fix is bridging that gap between pulling ideas from machine learning research, building a proof of concept and then making both the code open, and the learnt parameters of those machine learning models open, so that existing forecasting companies can take that and integrate it into their own stack.

"In order to have that relationship, you need trust. Being non-profit sends that loud signal that we're not competing with you, we're here to help you. It also helps us build a trusting relationship with academics and researchers and data providers as well."

It's a compelling argument, particularly given the early days of the project, but it's easy to envision ways that private industry - and in particular, the immense data power of the technology giants - could be harnessed for the increased climate impact that Subak wants to deliver. Travers is optimistic about the potential involvement of commercial organisations.

"The data cooperative we're forming is open for everyone, so that includes private companies, publicly listed companies [and] commercial enterprises. Part of the point of this is trying to bridge this gap, and we've got people coming from the commercial world as well as from the non-profit advocacy world."

Travers and Kelly both believe that organisations of all sorts are more willing to share important data relating to climate change now than they ever have been - although how much of this is motivated by altruism is debatable. Travers suggests that many organisations have moved from a point where they knew their data had value, but not how much - so they held onto it - to a position where their concerns about climate change have escalated and if that data isn't doing anything useful, why not open it up so it can?

Enterprises are also becoming increasingly conscious of their ESG profiles. It looks good if they are seen to be working with the non-profit sector to put something back. At the same time, regulators like Ofgem are signalling that they would also like data to be made open, and are exerting pressure.

From hours to minutes

Machine learning is a hot topic, and the question of what AI can contribute to the fight against climate change is one that enthuses Kelly - although he is keen to emphasise that it is no magic bullet.

"AI is just one of a whole suite of tools that we have. The fact that you can deploy it without having to build billions of pounds worth of physical infrastructure is helpful."

Much of Kelly's work involves ML research, and Open Climate Fix's specialty is in applying these concepts to cloud movements. The problem of trying to predict where clouds are going to move over the next few hours is a big one when it comes to forecasting solar electricity generation (and for reasons that would take more space than we have, can save the Grid a lot of time and power). This near-term forecasting is an area where AI can really make an impact.

"It takes a while to train, but then once it's trained you might get a forecast in a minute or two, whereas conventional weather forecasting models take a few hours. For things like clouds, which change very quickly, being able to generate a forecast really quickly after last observing the state of the world is really important."

Technology in itself will not save us from climate change, but the more data is shared and the more easily the huge datasets that ML models need become available to any organisation or individual who can develop them, the closer net zero looks. Our future depends on it.

Subak is recruiting now - please click here for details.