Fighting Climate Change with AI: Just more ‘Jam Tomorrow’?

Fighting Climate Change Rob Bennett Image DALL-E — Made with DALL·E 2 httpss://openai.com/dall-e-2/

We’ve seen the defence and security community[1][2][3] say that they want to use AI to help them predict and direct their climate change actions[4][5][6]. And so they should – we’re running out of time to make a difference: the UK has just over 300 months to reach Net Zero[7] (or just 96 months, depending on who’s in charge).

To help make this happen, organisations have come together to specifically focus on AI and climate change[8][9][10], and several businesses have appeared with new climate-focused AI products and services[11]. Both governments[12][13] and organisations[14][15] across the world have indeed stated the need to make both better decisions and faster decisions. In the UK, the Ministry of Defence’s Climate Change and Sustainability Strategy[16] looks to embrace the ‘fast follower’ concept of emerging technology. And tech and industry-specific bodies like techUK[17], ADS[18], and TDI[19] have been helping to navigate a way forward.

But here’s the rub: even with all those plans and all that help, too many traditional AI/ML solutions[20] rely on a volume, availability, and quality[21][22] of data that (when it comes to climate change), few organisations currently have[23][24][25]. True, some types of industry may be drowning in data on climate change, and other parts of the defence and security community indeed do capture and generate terabytes of other data[26][27], but many are simply starved of data andinsights[28]. While AI solutions promise to unlock the value of ‘big data,’ for most organisations, their data just isn’t ‘big’ enough [29][30][31].

This has resulted in those organisations finding themselves in a seemingly intractable paradox[32]: not knowing what data to prioritise, capture and clean without a solid climate change decision model[33][34], but also not being able to generate a decision model that helps them prioritise and invest without that very data.

Even if those organisations work with (or better still, invest in) data scientists[35][36] to decide up-front on what data to prioritise, they then must spend big on data management[37][38] and wait months or even years[39][40] before the data is ‘ready’ to be used by an AI solution. In fact, a large proportion of those AI solutions are really just adding more sensors to organisations, using publicly available data (not specific), automating the sifting and sorting of an organisation’s existing data, or enabling the reporting of it. All great stuff, but the deep and rapid insight – the ROI(C)[41] – that we truly need from AI supposedly still come (sometime) later.

On top of that, the inherent complexity and uncertainty of reality, compared to the often-simplistic data captured and hard-to-understand AI models produced[42], all combine to result in AI for climate change being nothing more than ‘jam tomorrow’[43]. Jam tomorrow, even though the security community wants climate change understanding, consensus, and (most importantly) action – right now.

This isn’t to say government climate change reporting requirements[44] aren’t supported by guides that help make progress[45][46][47]. And there are, of course, consultancies that specialise in climate change services[48][49] and some solid frameworks for AI in climate change[50]. Nor is this to say that there aren’t organisations and businesses that are leading from the front[51][52]. But any way you look at it, when it comes to AI, too many of the actual AI solutions on offer make things both expensive and slow.

The way out of the fog? Those organisations need to adopt a new approach that can embrace uncertainty[53][54][55] by using the messy and patchy data we already have straight away. And use then that data to build models in the face of low data and high uncertainty[56] that help them make explainable, defensible, and actionable[57][58] climate change decisions immediately.

There are, of course, other concerns about using AI[59] to solve our climate problems – from privacy and bias to the environmental impact AI itself causes. But if we are going to use AI to solve the issues in the defence and security community, we need the ‘AI jam’ for climate change today.

References:

[1] httpss://www.turing.ac.uk/partnering-turing/current-partnerships-and-collaborations/gchq-ministry-defence

[2] httpss://www.gov.uk/government/publications/national-ai-strategy-ai-action-plan/national-ai-strategy-ai-action-plan

[3] httpss://media.defense.gov/2021/Oct/07/2002869699/-1/-1/0/DEPARTMENT-OF-DEFENSE-CLIMATE-ADAPTATION-PLAN-2.pdf

[4] httpss://www.bcg.com/publications/2022/how-ai-can-help-climate-change

[5] httpss://www.forbes.com/sites/markminevich/2022/07/08/how-to-fight-climate-change-using-ai/?sh=4049d4292a83

[6] httpss://www.turing.ac.uk/blog/cop26-and-beyond-crucial-role-ai-tackling-climate-change

[7] httpss://www.gov.uk/government/publications/net-zero-strategy

[8] httpss://www.aifortheplanet.org/en/content/ai-for-the-planet-alliance

[9] httpss://www.climatechange.ai/

[10] httpss://impactlab.org/

[11] httpss://www.forbes.com/sites/markminevich/2021/10/08/11-examples-of-ai-climate-change-solutions-for-zero-carbon/?sh=5b49729f2251