What can a neuroscientist do about climate change?

The climate crisis is a defining challenge of the 21st century. If we keep on the same emissions path, we expect a sea rise of 2 meters, hundreds of millions of climate refugees, a mass extinction of over half of animal species, and decreases agricultural productivity that could leave a billion people starving by 2100. The IPCC report shows 4 paths of carbon reduction that keep warming to 1.5 degrees, each requiring massive changes in personal behaviours, political structure, economy and technology. Every sector of society will need to join the effort.

Neuroscientists are taking notice. The Society for Neuroscience (SFN) recently hosted a satellite event about tackling climate change – more than 80 people participated. Jeremy Freeman stepped down from CZI to focus on climate change; Adam Aron at UCSD published an editorial on why climate change needs attention from cognitive neuroscientists; and people like Konrad Kording, Demis Hassabis and Adam Marblestone are publishing detailed breakdowns of the climate crisis.

So what can a neuroscientist do about climate change? Neuroscience is not an engineering discipline, so it’s easy to feel powerless, but I’m going to try to show here we have a surprising amount of agency: personal agency, the power to convince, and the power to invent.

Personal agency

The SFN event ended in a petition to decarbonize the conference. This includes suggestions such as:

  • hosting satellite events closer to home to limit the amount of travel
  • skipping the conference every other year
  • integrating teleconferencing to attend remotely

As academics, much of our carbon footprint stems from flight. Assuming that the 30,000 attendees travel on average 10,000 kilometers to get to the conference, it’s roughly 70,000 tons of carbon in the atmosphere from air travel only – or the average annual CO2 emitted by 10,000 EU citizens. In research universities, air travel accounts for a large proportion of the carbon footprint of the institutions; for example, it’s estimated that over 50% of the carbon footprint of Université de Montréal is from air travel. There are already excellent models for zero-carbon conferences. Traditional international conferences don’t make sense in a carbon-constrained world, and we can refuse to participate in them.

Flying is one of most carbon-intensive behaviours scientists participate in, and one of the highest leverage items. Neuroscientists also consume resources to run their labs, and these resources – and the embodied carbon – are wasted when the products of the research are not shared. This is a strong argument for open science. A salient example of resource use is training state-of-the-art deep learning models, trading GPU cycles for classification accuracy. One recent analysis estimated a carbon footprint of half a ton of CO2 for training a single NLP model (BERT); an architecture search over different networks yielded numbers in the hundreds of thousands of tons. Computational neuroscientists should share their trained models to prevent the wasted resources from re-training. Experimentalists should also share their data for the same reason. Data carries an embodied carbon footprint from the materials used to collect it (plastics, animal facilities, the carbon used to ship materials across borders, etc.), and publishing the existing data makes more use of the same amount of resources.

Although I’ve focused on activities specific to scientists, we can participate in the same carbon-cutting behaviours recommended for everyone:

  • having fewer children
  • living car-less
  • eating a plant-based diet
  • using green energy
  • living in more modest houses
From Wynes and Nicolas (2017)

The power to convince

About 70% of people in the US believe that climate change is happening; that leaves 30% of people that remain to be convinced. Furthermore, only a small fraction of believers act. Neuroscientists, as teachers and science communicators, can help people graduate from disbelief to disengagement to action.

Scientists have a lot of influence. They are consistently ranked among the most trusted professions in the US, and they can bring attention to important matters via teaching, their online presence and their political influence. As Adam Aron points out in his editorial, people are more convinced about the urgency of the climate crisis when communicators adopt themselves a low-carbon lifestyle. So scientists’ power to convince is gated by their own adoption of the behaviours they want to see in the world.

As academics, we have a higher standard of what it means to understand something, so before we go convincing, we have the duty to become well-informed. I think we can all take time to better understand the mechanisms of climate, the evidence for it, and its consequences. Plenty of resources are available to learn these materials; a very good starting point is Michael Mann’s class on edX. Those more inclined to learn about the solution space should read the late David Mackay’s book.

Given that we’ve changed our own behaviours and learned how climate change works, we need to find ways to amplify our own actions and convince others to act. This forum can be the classroom; Adam Aron’s class on climate change psychology gives us a template. This also has to include political activism; we can vote for parties that support carbon taxes and decarbonization; join grassroots advocacy groups; and support and participate in civil disobedience.

At the research level, Adam Aron argues that cognitive neuroscientists can help the climate crisis by studying the mechanisms by which people acquire beliefs about climate change, how they propagate through their social networks, what kinds of messages they best respond to, and what motivates them to act.

For example, a group at MILA is working on generating flooded images of people’s homes so they can experience for themselves what it will be like to live in with a 6 foot sea level rise. The hypothesis is that people are emotional decision makers and will respond best to emotionally-loaded messages. These kinds of approaches are ripe to be studied from both a neuroscientific decision-making perspective and an econometric efficacy perspective. What causes people to graduate for disbelievers to believers? What causes people to move from belief to action? And how many tons of CO2 could we keep in the ground if we convinced 1,000 people?

The power to invent

Many readers of this blog are technologists, engineers and programmers in addition to neuroscientists. Technological innovation will be crucial to limit climate change to 1.5 degrees. Software allows us to communicate climate change, to visualize it, predict it, plan for it, and mitigate it. Physical engineering will allow us to build the large-scale structures required to keep carbon in the ground. Bret Victor wrote an excellent blog post on how technologists can help climate change, and it is a very encouraging read. A  consortium of machine learning researchers, including Yoshua Bengio, Demis Hassabis and Konrad Kording wrote a sprawling 100-page paper with a companion website on how artificial intelligence can help climate change. By all means, if you’re a machine learning expert, pick a section in the paper labeled high leverage, and go work on it.

Engineering is a very small part of neuroscience today, but it will grow in near future. Mainstream (i.e. 10M+ units) consumer-grade BCIs will likely appear on the market in the next 10 years. Meanwhile, decarbonization scenarios that will keep warming to 1.5 degrees will require sustained efforts over the next 80 years. Neuroscience research that might be considered far out has the potential to have massive impact in the future.

Emissions scenarios leading to 1.5 degree warming

Warning: speculation, thought experiments, future episodes of Black Mirror ahead

There are opportunities at the growing convergence of AI and neuroscience. Suppose, for instance, that by studying the brain, we discover an algorithm that opens up a rich area of AI (e.g. reinforcement learning, multi-agent learning, few-shot learning, causal inference, etc.). Some of the applications of this algorithm could be used effectively to mitigate climate change. For instance, suppose one of the downstream applications is counterfactual reasoning in life-cycle analysis; then we could very easily optimize a supply chain to minimize the carbon footprint of a given product. In this case, we’re augmenting human reasoning using AI, itself inspired by human reasoning, with the end goal of tackling climate change.

Computers are cognitive enhancers – Steve Jobs called them bicycles for the mind. People also use drugs to make themselves smarter, caffeine being the most salient example. People working on climate change may choose – indeed, may feel a moral imperative – to enhance themselves cognitively. The enhancement of well people (whether with nootropics, psychedelics or empathogens) is a taboo subject, but neuroscientists working on these areas  enhance the human potential to solve problems. The same point can be made about invasive BCIs for cognitive enhancement. Supplementing our memory, or having immediate access to a library of all human knowledge, could make us radically smarter, and more effective at fighting climate change.

Finally, humans could eventually choose to radically decrease their carbon footprint by leaving their physical bodies and uploading themselves to the cloud. While the capacity to emulate a single human brain in real-time using simplified biophysical models is not here yet, it should be feasible within the next 15 years. The cost of that simulation with today’s efficiency would be around a tenth of the power envelope of the United States. However, efficiency follows a similar trend to Moore’s law, with a doubling in operations per second every 1.5 years or so. At this rate, we could have whole brain emulation at a carbon intensity similar to that of 2019 human by 2055 – at which point uploading becomes carbon efficient, and rapidly more so over time. The carbon-negative technologies introduced in the 2060s might thus include foregoing our physical bodies.

Conclusions

I’ve highlighted three ways in which neuroscientists can help tackle the climate crisis: by changing their personal habits, by convincing others, and by inventing new ways to a carbonless future. We have to soberly evaluate our actions with respect to their causal impact, and live our lives accordingly. Neuroscience, like every sector of society, will have to change to tackle the defining challenge of this century.


Leave a comment

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s