The Atomic Human

edit

The Attention Reinvestment Cycle

The recent Draghi report on The Future of European Competitiveness bemoans the innovation gap between Europe,the US and China. The report looks how the gap in productivity between the US and Europe has widened since the 1990s. Draghi suggests that this is because Europe has missed out on the digital economy.

The Draghi report considers how Europe’s competitiveness needs to be improved across the next wave of the digital revolution. A motivation for improved productivity is economic growth which should lead to greater prosperity and political stability. But comparing to the US we see that despite their greater productivity, they face political instability and increasing insecurity about their global role. They are returning to isolationist debates not heard since before the second world war.

My ex-colleague at Amazon, Narayan Venkatasubramanyan, is a supply chain guru. He likes stories. At a Chinese restaurant in Palo Alto, he once told me the story of the last 150 years in supply chain. He explained that across the nineteenth and first half of the 20th century, manufactured goods were in short supply. If you could make it people would buy it. Those that controlled the means of production made large profits. Manufacturing was the bottleneck in supply chain and therefore the source of economic value.

After the second world war, manufacturing capacity increased. This meant the bottleneck shifted from our ability to produce to our ability to consume. Profits in the supply chain shifted from controlling supply to controlling demand.

A simplistic economic model would suggest that demand is controlled by the best value proposition. But in practice there is a further bottleneck. One of the foundations of The Atomic Human is the fact that human attention in limited.1 When we have an abundance of information, our attention becomes the bottleneck.

Our limited communication ability lies at the core of the atomic human. And our intelligence (both individual and cultural) has evolved ways through which we are selective in choosing which information to focus on. Because the new bottleneck is consumption, and the route to that consumption is through our attention, companies are economically incentivised to capture our attention. This phenomenon is known as the “attention economy”.2 The challenge of capturing attention is what the American department store magnate, John Wanamaker, was referring to when he says “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.”3 Today advertisers no longer face Wanamaker’s quandary. Because they can directly quantify how their advertising is performing.4 The emergence of System Zero5 is a consequence of this.

This is leading to what I call the “attention capture cycle”. Businesses are incentivised to make money by capturing our attention, the ones that do this well make more money and scale up dominating the attention economy. This takes us back to the Draghi report and a fundamental challenge that it doesn’t address. The attention capture cycle leads to a form of digital feudalism6 where the attention economy is dominated by a few large companies. This undermines our affordances both for our immediate wants and our long term aspirations.7

In a recent Financial Times Opinion piece, I highlighted that the challenge we face is that as we automate more, our human aspirations become squeezed by the necessity to quantify all aspects of the automated supply chain. This explains why in our public dialogues we receive a clear message about the type of technological innovation people would like to see. But what is supplied is entirely the opposite of what is requested.8 People request advances in education, healthcare, social care, security. They want to see progress on the so-called wicked problems. It seems such progress will require a more sophisticated approach to innovation than the current attention economy is allowing for.

So what is the solution? The Draghi report argues for a range of interventions including increased innovation funding, better translation of research into practice, and new attitude to regulation. Many of these ideas are also reflected in our own policy work within the UK. But we will need to do more than that.

I believe that as academics and universities, we need to do a better job of stepping up. Data Science Africa, Accelerate Science, the Data Trusts Initative and ai@cam represent a range of initiatives that involve collaborations across government, industry and academia. Each as a different purpose, but there is a shared philosophy to all the approaches. It is an attention reinvestment cycle.

Each initiative shares at least some of the following principles:

  1. Agility. Keep the size of the institutional infrastructure low so that it can adapt to evolving circumstances.
  2. Share credit. It’s amazing what you can achieve if you don’t care who gets the credit.
  3. Scaling Human Capital. Encourage reinvestment of your community’stime, scale by having participants see one, teach one, organise one.
  4. Bidirectional learning. Wisdom comes from diversity, as technologists we learn as much from the domains as they learn from us.

Alongside these principles, there is a final ingredient that enables an attention reinvestment cycle.9 Chapter 11 of The Atomic Human describes the notion of the Human Analogue Machine10. With the advent of generative AI, for the first time we have an interface to the machine where instructions can be given in natural language. This means that when we work closely with domain experts, understand their role, and guide them to the right solutions for making their role easier we free up their time and create the potential for a virtuous cycle of attention reinvestment. Once they have understood how to deploy the tools, they become the expert who is best placed to share their knowledge with others. Applied at scale, this would mean that the answer to the question “how” is not being given by “technical experts” but by nurses, doctors, teachers, local planning officers.11

The attention reinvestment cycle.

The attention reinvestment cycle leverages the principles that have emerged from the initiatives we’ve reviewed to reinvest time savings into the wider sharing of how those time savings were made. This refocuses the innovation economy. By reinvesting time directly, rather than returning financial savings, we can ensure that the innovation economy focusses on the challenges that those individuals are prioritising, rather than the attention-capture priorities of our current system.

In the United States, the geopolitics of AI is dominating decision-making. As insecurity about their power increases, big tech companies are seen as an important instrument in defending US power. But the choice between corporate autocracy and government autocracy is a poor one.12 As the Draghi report emphasises, the challenge for Europe (and the UK) is to bring the benefits of these technologies in ways that reflect our values. Values that can be traced to Popper’s defence of the Open Society. The attention reinvestment cycle offers a route forward to achieve this aim.

Machine Commentary

NL: How do the themes in this reflection relate to the book The Atomic Human?

Claude: This reflection connects to several key themes from The Atomic Human:

  1. Limited Human Attention:
    • The reflection builds on the book’s core premise about human attention being limited (Chapter 1: Gods and Robots)
    • References the “atomic human essence” concept directly in discussing our limited communication ability
    • Connects to the book’s discussion of human bandwidth constraints and the “embodiment factor”
  2. System Zero and Digital Oligarchy:
    • Expands on the book’s concept of System Zero (Chapter 8) in describing the attention capture cycle
    • Develops the book’s warnings about digital oligarchy and corporate control
    • Shows how attention economy leads to concentration of power discussed in the book
  3. Human-Analogue Machines:
    • Builds on Chapter 11’s discussion of the Human Analogue Machine concept
    • Shows how natural language interfaces create new opportunities for human-AI interaction
    • Emphasizes the book’s point about maintaining human agency
  4. Cultural Evolution and The Open Society:
    • Connects to the book’s emphasis on cultural context in intelligence
    • Develops the book’s argument about preserving democratic values in technological development
    • Shows how European/UK approach could differ from US model
  5. Wicked Problems:
    • Reflects the book’s epilogue call for practical institutional responses
    • Shows how initiatives like Data Science Africa and Accelerate Science implement book’s principles
    • Demonstrates the “piecemeal social engineering” approach the book advocates

NL: How does this reflection build on the framework laid out in the 5 Ps model?

Claude: The attention reinvestment cycle directly builds on and exemplifies the 5 Ps framework:

  1. Purpose:
    • Clear articulation of purpose: redirecting innovation toward public benefit while preserving democratic values
    • Addresses the gap between what people want from technology (education, healthcare, social care) and what is currently delivered
    • Purpose emerges from societal needs rather than being imposed top-down
  2. People:
    • Emphasizes the importance of domain experts (nurses, doctors, teachers, planning officers)
    • Highlights the need for diverse voices and bidirectional learning
    • Shows how people convene around shared purpose in initiatives like Data Science Africa
  3. Projects:
    • References concrete initiatives (Data Science Africa, Accelerate Science, Data Trusts Initiative, ai@cam)
    • Shows how projects emerge from understanding domain expert needs
    • Demonstrates how projects can scale through “see one, teach one, organize one” approach
  4. Principles:
    • Explicitly lists four key principles:
  5. Agility in institutional infrastructure
  6. Credit sharing
  7. Scaling human capital
  8. Bidirectional learning
    • These principles emerge from successful project experiences
  9. Process:
    • Describes a sustainable process for reinvesting attention
    • Shows how time savings can be reinvested into sharing knowledge
    • Demonstrates how process grows organically from principles and projects

The reflection shows how the 5 Ps framework can be applied to create a virtuous cycle that addresses the challenges of attention capture while building sustainable institutions.

NL: How do the learnings from the four initiatives mentioned inform the approach described in this post?

Claude: The four initiatives (Data Science Africa, Accelerate Science, Data Trusts Initiative, and ai@cam) each contribute key learnings that shape the attention reinvestment cycle:

  1. Data Science Africa’s Contributions:
    • Demonstrates successful scaling through “see one, teach one, organize one” principle
    • Shows how local expertise can be amplified through knowledge sharing
    • Proves the effectiveness of keeping institutional infrastructure light and agile
  2. Accelerate Science’s Learnings:
    • Illustrates how to bridge between technical experts and domain specialists
    • Shows how to maintain focus on practical outcomes while building capability
    • Demonstrates the value of bidirectional learning between technologists and scientists
  3. Data Trusts Initiative’s Insights:
    • Provides framework for institutional governance that preserves democratic values
    • Shows how to build trust while enabling innovation
    • Demonstrates importance of bottom-up approaches to data stewardship
  4. ai@cam’s Experience:
    • Shows how to coordinate across diverse stakeholders while maintaining agility
    • Demonstrates university’s role as “honest broker” in technological development
    • Illustrates how to balance innovation with public benefit

Common Patterns Across Initiatives:

  1. All emphasize empowering domain experts rather than replacing them
  2. All maintain light institutional structures while achieving significant impact
  3. All focus on practical outcomes rather than theoretical frameworks
  4. All demonstrate the importance of bidirectional learning
  5. All show how attention savings can be reinvested productively

These learnings directly inform the attention reinvestment cycle by showing how time saved through automation can be effectively channeled into knowledge sharing and capability building while preserving human agency and democratic values.

  1. As we reviewed in spiritual animals, see also a recent paper in Neuron that talks about slow speed of thought, but I think the paper misses the fact that this is likely determined by our limited speed of bidirecitonal communication. Their estimate is 10 bits per second. They use similar techniques to the Atomic Human, only the estimate in the Atomic Human is based on Shannon’s estimate of informaton in language and the number I use is around 30 bits per second. 

  2. See Spiritual Animals for more detail on this. 

  3. As referred to in The Cultural Conundrum

  4. I first commented on this emerging phenomenon for an article in the Guardian from July 2015. You can see a commentary on this article here

  5. Covered in Chapter 8 of The Atomic Human

  6. See A Retrospective on Digital Oligarchy for more on this. 

  7. See Bridging from Domain Experts to AI Capability for more on this. 

  8. See Art is Human and this retrospective on Royal Society Public Dialogue

  9. See also my talk for the Bennet Public Policy Lecture 2024 here and with my notes here

  10. See The Human Analogue Machine for more on this. 

  11. This is what Karl Popper refers to as the “piecemeal social engineers” in “The Open Society and its Enemies”

  12. The question of this choice was put to me in an interview I did for BBCs Hardtalk. If you’re in the UK you can watch the interview here otherwise it’s available in podcast form here

Click to see what the machine says about the reflection and the book