Three Big Ideas #27
Fixing the forecasters, token logic, and siting smart
Welcome to our weekly Three Big Ideas roundup, in which we serve up a curated selection of ideas (and our takes on them) in entrepreneurship, innovation, science and technology, handpicked by the team.
🧮 Eamonn Ives, Research Director
Some norms in British politics – like elections always falling on a Thursday – have been around for generations. Others sometimes feel like they have been, but are actually very recent phenomena. The sacrosanct involvement of the Office for Budget Responsibility in Treasury decision-making is one such example. Despite only being set up in 2010, the OBR enjoys an extraordinarily privileged position, and wields immense power over the government of the day.
We saw that on full display last week when Rachel Reeves delivered her Spring Statement, and made painstakingly exact spending promises in order to meet her fiscal rules. After all was said and done, the Chancellor maintained precisely the same headroom – £9.93 billion, down to two decimal places – as she enjoyed after her October Budget last year.
A defender of the OBR would argue that its scrutiny keeps chancellors honest and Britain’s reputation in the markets sound (or at least, more sound than the case would be otherwise). But an interesting challenge from economists Pedro Serôdio and Rohan Shah which recently caught my eye argued that the OBR is falling short in certain respects, especially when it comes to forecasting the impact of medium- and long-term policy changes.
This is enormously important for the future of our economy. Some of the things that matter most for increasing growth are reforms that the OBR, in their analysis, is not well set up to model. Think here of changes to planning frameworks that could deepen labour markets by allowing more people to live in the most productive parts of the country.
As well as convincingly diagnosing the issue, Serôdio and Shah prescribe a solution. For very little investment in the grand scheme of things, they outline how to bolster the OBR’s talent pipeline and enable it to do a better job of forecasting policy changes, and thus strengthen the scope policymakers have to make meaningful reforms. I won’t try to forecast the benefit of doing so myself, but I’d bet the answer would be net positive.
🧩 Anastasia Bektimirova, Head of Science and Technology
Think about how we’ve traditionally organised our economy. We group similar activities into industries, divide work into job categories, and create companies around related products and services. These divisions make intuitive sense to us as humans. But AI systems don’t share our intuitions.
AI systems divide information into ‘tokens’ – discrete units based on statistical patterns, rather than human meaning. Unlike human categories based on meaning and function, tokenisation creates divisions based on what works computationally rather than what makes intuitive sense. As Nicklas Lundblad, DeepMind’s Senior Director of Policy and Strategic Advisor, writes in his blog, the tokenisation process is already reshaping major industries:
“Consider how digital platforms have restructured industries by tokenizing previously continuous experiences. Uber tokenized transportation into discrete, algorithmic units; Airbnb did the same for accommodation; TikTok for entertainment. Each platform succeeded by reconceptualizing an industry in terms that could be discretized, quantified, and optimized according to computational logic. I think this comes close to what Andreesen meant when he noted that software will ‘eat’ everything – but in order to do that it first needs to digest the world, into tokens.”
As tokenisation spreads, our traditional industry boundaries might blur in unpredictable ways. We can already see this in how companies like Amazon and Apple operate across what were once distinct sectors, following patterns that transcend standard categories.
What I find most interesting about this line of thought is implications for institutions. Our institutions are organised around human-meaningful categories that emerged in specific historical contexts. If tokenisation fundamentally reorganises economic activities, then institutions – which are “the humanly devised constraints that shape human interaction” – may misalign with economic reality.
New technologies have often demanded new institutional forms – for example, the industrial revolution created corporations and regulatory agencies. Similarly, institutions will likely need to be designed around the actual patterns of a tokenised economy – potentially cutting across traditional domains like healthcare, finance, and education to address newly visible patterns of risk and opportunity. The greatest challenge ahead may not be the technology itself, but reimagining our institutions for a world where the organisation of economic activity no longer follows the patterns we’ve built our social structures around.
🛰️ Jessie May Green, Events and APPG for Entrepreneurship Coordinator
You may have heard a lot about the clean energy transition, but have you ever wondered what it looks like? With the Global Renewables Watch – a partnership between The Nature Conservancy, Planet Labs and Microsoft’s AI for Good Lab – you need wonder no more. Together, they have mapped the spread of onshore wind and large-scale solar over time using advanced AI and high-resolution satellite imagery.
In this interactive article by The New York Times, you can visually explore how renewable energy capacity has grown over the last eight years. For instance, the US’ solar and wind capacity has nearly tripled, China has built more than 120,000 wind turbines – almost a third of the world’s total, and emerging economies like Turkey are beginning to fulfil their solar potential.
Despite this progress, still around three quarters of global greenhouse gas emissions are generated by energy use, and so cleaner energy infrastructure must continue to be rapidly scaled to mitigate further global warming. This will require a lot of land. If not done thoughtfully, the expansion of renewables could – ironically – cause a lot of environmental damage, not to mention human conflict.
That’s where ‘smart siting’ can come in – choosing wind and solar sites based on where they’ll have the fewest negative impacts. The Global Renewables Watch dataset is unique in that it captures trends over a period of time, not just snapshot moments, and tracks underlying development patterns, not just development. Thus, it can predict where renewables siting may cause tensions to flare up, and assist in producing alternative plans. Considering how conflict can impede action on climate change, this is no small matter.
With Microsoft providing the AI and platform technology, Planet Labs on the satellite imagery, and The Nature Conservancy bringing the expertise needed to analyse the trends, this is a stellar example of the importance of cross-sector collaboration on climate action. With all the doom and gloom reports about AI threatening the environment, it’s nice to remember the positives too.
 | A guest post by
|