To say that it’s the big question, the only real question for the long-term future, would be, in a way, an understatement. Climate change is happening, and it’s tremendously frightening for everyone. There are no silver bullets, no get out of jail free cards, and no exemptions for dealing with what’s happening to the planet.
If you go back and read my blog posts from February or March, we had a lot on the convergence of AI and climate policy. When I say “climate policy”, I’m also including a lot of grassroots efforts and business initiatives, kind of the whole ball of wax, as we all try to work together to address the problem.
Recently, we had some interesting input from Jaime Oliver, and I want to go over this, because I had really never heard things explained quite in this way before.
Oliver works at Clarity.AI, where researchers are working to make energy and ecology data more transparent in the business world. From the company’s web site:
“Whether you need a comprehensive, customizable, fully-packaged Sustainability Tech Platform or just one data point to ensure regulatory compliance, Clarity AI empowers you to efficiently and confidently assess, analyze and report on anything valuable to you or your clients and everything required by regulation.”
So Oliver was telling us how all of this works, and it’s interesting, because of the information models and techniques they work on.
Europe Illustration High-Res Vector Graphic – Getty Images
“Towards the end of the century, the planet is going to warm by 2.7 degrees,” he said. “According to … scientists, this is well beyond several climate tipping points. And so we need something more than just policies to bring sustainability to work. One particular way that people have thought about (this) is basically integrating sustainability into financial markets. So the main framework of what you call sustainable finance is that investors would divest from companies that don’t align with their values. And so they provide the right incentives for them to change behavior.”
To do this, he said, companies need the right input, and that leads them to ESG ratings. But, Oliver suggested, the ratings themselves lack the robustness and accuracy to really do the job.
“The data underlying these readings is not quality enough, and then it doesn’t have enough coverage of companies,” he said, describing Clarity.AI’s approach a bit. “We’re going through our three different pillars to add value to these data sets, and then serve it to clients through different methodologies. … we go to data collection … gather all the information from corporate sustainability reports, information that the companies are disclosing, the information that is not available, we complete it through AI models, and then we curate the different human errors that are kind of introducing the process.”
He explained how LLMs examine sustainability documents, news articles and earning calls to put together a better picture of a sustainability landscape.
“We solve scalability of the data collection,” he added, “but there are many data points that are not out there, because the companies don’t just don’t disclose them … so what we would do is, basically, complete that data set. What we do is basically learn from the companies that are disclosing this information, and generalize to the companies that are not disclosing. … We do this through machine learning models. They are not genAI models. So they’re not cool. I know. But, they work. And they capture rich interactions between the different features of a company, which are kind of superior to previous models.”
So in a nutshell, they are taking the available data, and extrapolating it to cover the “dark side,” and according to Oliver, it works!
“We were actually so sure about this, that we went to the very same researchers that were trashing ESG rating companies, and we proposed a collaboration with them. And that is actually what we did last fall here at MIT, where we were studying the co2 emissions that companies disclose, and the information that is available out there. What we found was that companies that went through a third party assurance process were reporting almost 10% higher emissions than their peers.”
That, he said, speaks to the lower quality of available data, but it also indicates that the companies with the third party systems in place have good intentions, as, according to his research, they tend to decarbonize faster than others.
“In the near future, what we want to do is scale the adoption of these datasets on these methodologies,” he said. “And basically, what we’re doing now is building what we call the ClarityAI assistant. With this assistant, (a genAI model,) through natural language, you can basically access ClarityAI database methodologies and answer questions around … those methodologies, how they work.”
That seems to be the general value proposition: that you can use the resulting data to fine-tune ESG, drive divestment (or other optimization) and promote real results in tackling climate change.
It’s worth a try: we have to do what we can to try to keep the planet’s temperature as low as possible. Look for more on this in the future.
