Announcing Climate Companion

Bringing…people in—with their networks of influence, their knowledge, and their resources—is the key to creating the capacity for shared intelligence that we need to solve the problems we face, before it’s too late. Our goal must be to find a new way of unleashing our collective intelligence…

Al Gore, The Assault on Reason: Our Information Ecosystem, from the Age of Print to the Era of Trump. Bloomsbury Publishing, 2017

My awareness of the dangers of climate change and my intention to act started when I saw An Inconvenient Truth, Thus, quoting Al Gore seems perfect today, as I announce the next chapter in my journey to reverse global warming. I’ve been talking for many years with friends and associates about what we can do about climate change. Then about 19 months ago, I quit my job to focus on contributing to climate change solutions, and started this blog in January.

Several months ago, my message about climate change started really hitting home with my friend and prior collaborator Darla Hewett. Darla is an amazing data engineer who has helped practitioners in multiple industries increase their effectiveness. Some of the technologies she uses resonated with and have been inspired by Doug Engelbart.

After some discussion, Darla and I decided to start a new initiative together, focusing on solutions to climate change. This post tells our origin story – read on the learn more.

Food Waste?

As I discussed in a prior post, reducing food waste has high potential to help reverse global warming. Food waste is a problem that both Darla and I encounter periodically, whether it’s a child not liking a meal, or the strawberries forgotten in the back of the fridge. So when we were deciding where to start with our initiative, food waste was compelling.

Next, we brainstormed on ways to attack this problem. We looked at general solutions to climate change, and specific approaches to food waste reduction. We divided the general solutions into seven categories:

1. R&D
2. Policy
3. Infrastructure
4. Decision support
5. Education
6. Funding
7. Tracking 

As we researched these, we discovered the Refed nonprofit, which analyzes 27 opportunities to reduce food waste in the United States. We crunched some of their numbers, and some of our own, to decide how to focus our food waste efforts within our seven categories. We found that decision support and tracking fit for our goals most closely. We decided to build an app to help the home consumer reduce their food waste. 

But, before we started building the app, we analyzed its potential greenhouse gas impact. We looked at research from the FAO and Drawdown. We made some assumptions about the market share that we could capture and the amount of food waste we could eliminate. We crunched the numbers. The answer? We estimated we could save about 0.01-0.02 Gigatons of CO_2e greenhouse gas emissions per year.  

This may sound like a lot, but we were disappointed. It wasn’t as much impact as we had been hoping. We wanted to get closer to half a gigaton of greenhouse gas reductions: this is the climate impact threshold used by Breakthrough Energy Ventures. 

We backed up and brainstormed. How could we get where we wanted to go?

An “Aha!” Moment

We thought about the climate crisis, about the need for multiple diverse solutions. The need to empower individuals and communities to contribute. The need for continuous improvement. How could we leverage our skills and knowledge to make a bigger impact and meet these needs?

We looked at our background and experience. I have a PhD in machine learning and natural language processing, and a diverse career applying AI in industry, and teaching it in academia. I’ve created adaptive conversational interfaces and led data science teams. Darla is a serial entrepreneur, a data engineer who invented a graph database approach and applied it in multiple industries. She worked with Doug Engelbart and is inspired by his vision of collective intelligence.

We realized that we could combine forces to create something that few others could. What if we could enable multiple solutions, that together could lead to the half a gigaton of impact we wanted?

Impact Platform

In sum, our insight was that we could enable multiple solutions by building not just an app, but a platform. We still stood by our hypothesis that decision support will be key in addressing climate change. Looking at our toolkits, we saw that we could combine parts of them to build a decision support impact platform. The key pieces are Artificial Intelligence (AI), a graph knowledge base, natural language interaction, and tracking. 

First, AI provides recommendation capabilities that assess options and preferences and rank actions accordingly. We also incorporate machine learning for classifying inputs. For example, we will want to classify food images when reducing food waste. Machine learning also supports clustering to recognize patterns, such as common recipe characteristics. Similarly, time series models help us forecast outcomes such as food consumption over time. Finally, decision networks can be used to weigh options automatically, incorporating constraints appropriately.

Next, we have a novel and powerful capability to represent complex knowledge using graph-based structures. In our approach, graph links are first-class objects, which enables multiple viewpoints to be represented by the same underlying structure. We call this capability our universal knowledge representation. In the food waste domain, it allows powerful search for recipes that minimize food waste while respecting dietary constraints.

Third, current technologies for natural language interaction support communication in text or voice as needed for the situation. Users can train their personal chatbot to interact in a way that is appropriate to their context. Together with the AI and knowledge representation, this lets all participants make progress independently of programmers.

Finally, tracking allows decision makers to see and validate the climate impact of their decisions over time. Communities can demonstrate impact to others who are interested in pursuing similar solutions. 

In sum, our decision support impact platform confronts the enormity and complexity of the climate challenge head on. But of course, being data driven, we analyzed the potential impact of our platform on climate change. Not surprisingly, there is high variability in our estimates since we’re uncertain how much market share we can capture. And the variability increases with each domain added. In the end, we estimated that we would need between 3 and 25 apps on our platform to reach the half gigaton threshold. Given that the development of each new app will be accelerated by having a platform, we decided that this is a great investment of our time!

And we’re still starting with food waste, to focus the development of the platform MVP.

Today on the blog, we’re announcing Climate Companion, and we are building out our vision now. In the long term, we plan to offer our platform to others working on climate change. Our goal is to bring in your solution development and knowledge contributions, support blending them with other solutions, and help you track your climate impact. With all hands on deck, the system can learn what works and doesn’t, and amplify all solutions for maximum impact. Together, we’re all Climate Restoration Amplified, using Climate Companion. Welcome to the next stage in the journey!