Do facts have an expiration date? (with Samuel Arbesman)
Samuel Arbesman, Scientist in Residence at Lux Capital, discusses how facts have a half-life, emphasizing science as a self-correcting process and the importance of epistemic humility. He advocates for rethinking technology as a humanistic liberal art to foster wonder, meaning, and bespoke creations.
Deep Dive Analysis
17 Topic Outline
The Concept of the Half-Life of Facts
Quantifying Knowledge Obsolescence in Scientific Fields
Distinguishing Fact Refinement from Overturning
How Measurement and Definitions Reshape Knowledge
The Challenge of Updating Meso-Scale Facts
Scientific Theories as Mental Models and Metaphors
Cultivating Epistemic Humility in Understanding
Science as a Self-Correcting Human Endeavor
Fundamental Elements for Effective Scientific Progress
Exploring New Models for Research Organizations
Reframing Computing as a Humanistic Liberal Art
Designing Technology for Wonder, Meaning, and Connection
AI's Potential for Democratizing Software Creation
Catalyzing the Adjacent Possible in Innovation
Innovation Through Replicating, Remixing, and Reusing
Understanding Why Technologies Sometimes Regress
Embracing Wonder and Excitement in Science
8 Key Concepts
Half-Life of Facts
This concept uses an analogy to radioactive decay, suggesting that while individual facts are unpredictable, a large body of scientific knowledge shows regular patterns in how it changes, becomes obsolete, or is overturned over time. It implies that scientific knowledge is not static but dynamically evolves.
Asymptotically Approaching the Truth
This idea posits that scientific knowledge, despite being in a constant 'draft form' and subject to revision or overturning, is continuously getting closer to a true understanding of the world. It frames scientific progress as an ongoing process of refinement towards an ultimate truth.
Meso-Facts
This category refers to knowledge that changes on the order of decades or a human lifetime. These facts are particularly challenging to update because they are often learned alongside more stable, unchanging information, leading individuals to forget to revise them as they evolve.
Epistemic Humility
This is the attitude of recognizing that our understanding of the world, including scientific knowledge and complex technological systems, is often incomplete and provisional. It encourages acknowledging the limits of human knowledge while still striving for better comprehension, rather than demanding perfect understanding or succumbing to complete ignorance.
T-shaped Individual
This describes a person who possesses deep expertise in a specific discipline (the vertical bar of the 'T') while also being comfortable and capable of crossing different domains and fields (the horizontal bar). Such individuals are valuable for bridging disciplinary gaps and fostering interdisciplinary connections in research.
Computing as a Liberal Art
This perspective views computing and computation not merely as a technical branch of engineering, but as a humanistic field. It emphasizes how computing touches upon language, philosophy, biology, and art, serving as a powerful tool for thought and for enhancing human experience and creativity.
The Adjacent Possible
Originally developed by Stuart Kauffman, this concept describes the set of all things that are possible given the current state of the world, whether in terms of existing technologies, scientific knowledge, or ideas. 'Catalyzing' it means actively lowering the activation energy to discover and realize these potential possibilities.
Stepping Stones (in Innovation)
In the context of searching a high-dimensional space for innovation, this strategy suggests that instead of directly aiming for a specific end goal, progress is best made by optimizing for interestingness or novelty. The discoveries made (stepping stones) can then be productively recombined in surprising and unexpected ways to lead to further advancements.
13 Questions Answered
Yes, similar to radioactive materials, while individual facts are unpredictable, a large body of scientific knowledge exhibits regularities in how it changes, becomes obsolete, or is overturned over time.
Science progresses by asymptotically approaching the truth; even as knowledge is overturned or refined, the overall endeavor leads to a better and closer understanding of the nature of the world.
Shifting definitions (like Pluto's demotion) and improved measurement techniques (like counting chromosomes or measuring Mount Everest) often lead to differing or more refined understandings of the world, even if the underlying reality hasn't changed.
Knowledge that changes on the order of decades or a human lifetime (meso-facts) is particularly hard to update because it's often learned alongside more stable, unchanging facts, leading people to forget to revise it.
The right attitude is one of epistemic humility, recognizing that scientific knowledge is in draft form and our understanding is incomplete, but we are continuously moving towards better comprehension without needing perfect understanding or complete ignorance.
Trust in science should be placed in its self-correcting process and rigorous methods for understanding the world, rather than viewing it as an infallible body of 'big T truth,' acknowledging that it's a deeply human and imperfect endeavor.
Essential features include the ability to share results, provide enough information for reproducibility, build upon others' work, and have mechanisms for providing credit to incentivize crucial activities like replication and tool-building.
Traditional academic structures often incentivize a narrow subset of valuable scientific activities (e.g., tenure-track research); new organizations are needed to support interdisciplinary work, tool-building, and diverse career paths that don't fit neatly into existing departmental lines.
Computing, when viewed broadly, is a tool for thought that touches upon language, philosophy, biology, and art, enabling new ways of thinking and creating, rather than solely being about technical implementation.
By designing technology that helps us achieve our most fulfilled and meaningful selves, such as AI tools that stitch together ideas, physical computing that fosters communal interaction, or creative coding that unlocks new artistic expression, rather than just focusing on productivity or distraction.
While AI democratizes software generation, traditional coding knowledge remains valuable as it makes AI tools more powerful and provides a deeper understanding of computational thinking, which is a constantly evolving skill.
It means lowering the activation energy to discover and realize new inventions, technologies, and ideas that are currently possible given the existing state of knowledge and tools, often by optimizing for interestingness and novelty rather than direct goal-seeking.
Technological progress is not an inevitable wave but depends on societal choices, individual decisions, and sustainable business models; factors like legislative limitations, lack of economic viability, or loss of specialized expertise can lead to technologies falling out of common use.
24 Actionable Insights
1. Embrace Science as Draft
Adopt the mindset that scientific knowledge is always in draft form, constantly being updated and improved, rather than a fixed body of truth. This helps in accepting new discoveries that overturn previous understandings without despair.
2. Trust Science as Process
Shift your trust from science as a static body of knowledge to science as a dynamic, self-correcting process of rigorously querying and understanding the world. This allows for acceptance of its inherent messiness and constant evolution, rather than being disillusioned by errors.
3. Cultivate Epistemic Humility
Practice epistemic humility by acknowledging that we don’t know everything and that’s acceptable, especially when dealing with complex systems or scientific knowledge. This helps avoid the false dichotomy of perfect understanding or complete ignorance, fostering a more productive approach to learning.
4. Hold Mental Models Loosely
Accumulate many mental models but hold them loosely, especially when applying them outside their original domain. This prevents oversimplification or drawing dangerous, grand sweeping conclusions from powerful but metaphorical theories.
5. Balance Skepticism Wisely
Maintain a healthy dose of skepticism towards scientific claims, recognizing that science is a human endeavor with imperfections and self-correcting mechanisms. Avoid unhealthy skepticism that dismisses all scientific findings due to isolated errors or frauds, which can lead to science denialism.
6. Approach Truth Asymptotically
Recognize that knowledge acquisition is a process of asymptotically approaching the truth, meaning we continuously get closer to a true understanding even as previous facts are overturned. This perspective fosters a healthy and optimistic view of scientific progress, seeing change as refinement rather than invalidation.
7. Choose Wonder & Excitement
Cultivate a sense of wonder and delight regarding science and technology, choosing excitement over fear or despair, even amidst uncertainties and changes. This positive outlook is integral to scientific and technological advancement and personal engagement with new discoveries.
8. Build Desired Future
Instead of solely predicting the future, focus on envisioning the kind of world you want to live in and then actively work to make that world more likely through technological and societal choices. This shifts the approach from passive prediction to active creation and purpose-driven development.
9. Design Tech for Meaning
Intentionally design and use technology by first considering how it can make life more fulfilled or meaningful, then working backward to envision what those technologies would look like. This shifts the focus from mere efficiency or time-wasting to human well-being and purpose.
10. Computers: Tools for Thought
Remember that computers are tools meant to enhance human capabilities and make us the best versions of ourselves, rather than ends in themselves. This perspective helps design and use technology for wonder and fulfillment, not just efficiency or passive consumption.
11. Computing as Liberal Art
Reframe computing and computation not merely as engineering but as a humanistic liberal art that touches upon language, philosophy, biology, and art. This broader perspective can lead to a healthier approach to how technology engages with human experience.
12. Optimize for Interestingness
When exploring high-dimensional spaces for innovation or research, optimize for interestingness or novelty rather than directly pursuing a specific end goal. Treat discoveries as ‘stepping stones’ for productive recombination, leading to surprising and unexpected outcomes.
13. Catalyze Adjacent Possible
Act as a catalyst to lower the activation energy for exploring the ‘adjacent possible,’ making it easier to discover and actualize potential inventions, technologies, and ideas. This involves fostering environments where new combinations and discoveries are more likely to occur.
14. Connect People & Ideas
Act as ‘connective tissue’ by actively linking individuals and ideas across different disciplines, fostering dialogue and overcoming jargon barriers. This interdisciplinary approach is a powerful means of catalyzing innovation and generating new insights.
15. Cultivate T-Shaped Individuals
Foster individuals with deep expertise in one area (vertical bar of ‘T’) combined with the ability to jump across different domains (horizontal bar of ‘T’). This bridges disciplinary gaps and enriches research and problem-solving.
16. Experiment with Research Organizations
Explore and experiment with diverse institutional forms for research beyond traditional universities or corporate labs. This can incentivize valuable scientific activities not currently rewarded by existing structures, unlocking new kinds of discoveries.
17. Facilitate Research-Career Mobility
Create mechanisms that allow individuals to move fluidly between research and non-research roles (e.g., industry, startups) and back again. This prevents talent loss and enriches both domains by allowing diverse experiences to inform research and application.
18. Study Tech History
Study the history of technology to understand path dependence, contingencies, and past innovations. This knowledge can reveal ideas that were not ready in their time but can be repurposed or remixed for current innovation, providing powerful means for advancement.
19. Leverage AI for Idea Discovery
Utilize AI tools to stitch together disparate ideas and surface papers, articles, or concepts that would otherwise remain hidden due to jargon barriers or the vastness of available knowledge. This enhances intellectual exploration and understanding across fields.
20. Create Bespoke AI Outputs
Use AI tools to create highly personalized, ‘bespoke’ software, songs, or images tailored specifically to individual preferences and needs. This allows for the creation of unique digital experiences that resonate deeply with an audience of one, fostering personal fulfillment.
21. Explore Physical, Communal Computing
Investigate and develop physical and communal computing experiences, such as manipulating tangible objects on a table with projected interfaces. This can foster deeper human connection and tangible interaction, counteracting the isolating nature of screen-based computing.
22. Practice Creative Coding
Engage in creative coding to build artistic computer programs, exploring new forms of creativity that leverage the computer’s ability to perform vast calculations for generating unique images and animations. This offers a new window into art and expression previously unavailable.
23. Learn Traditional Coding
Continue to learn traditional coding alongside new AI tools, as deeper coding knowledge enhances the power and effectiveness of generative AI for building sophisticated software and overcoming limitations. This foundational understanding makes one a more successful user of AI tools.
24. Export Scientific Mindset
Work to convey the scientific mindset – viewing knowledge as a constantly improving draft – to the broader public. This can help foster a healthier understanding of scientific progress and its inherent uncertainties, reducing public anxiety about changing ‘facts’.
8 Key Quotes
Remember what I taught you? It's wrong. And if that bothers you, you need to get out of science.
Samuel Arbesman (quoting his professor)
If you think both of these things are equally wrong, then kind of your your your version of the world like your understanding the world is wronger than both them put together.
Samuel Arbesman (quoting Isaac Asimov)
All models are wrong, but like but some are useful.
Samuel Arbesman (quoting George Box)
We don't need to be in one of two states of like either like perfect understanding or complete ignorance.
Samuel Arbesman
We should trust science not as the body of knowledge but much more as the process of kind of trying to get better at understanding the world.
Samuel Arbesman
Computers aren't the thing, they're the thing that gets us to the thing.
Samuel Arbesman (quoting a character from 'Halt and Catch Fire')
An app can be a home-cooked meal.
Samuel Arbesman (quoting Robin Sloan)
Technological progress it's not this thing, it's not like this force or this thing you kind of pour over like innovation and suddenly we get we get new things, it's like it is something we have to choose to do as a society.
Samuel Arbesman