Notes on Engineering Health, July 2024: Notes on Unconventional Computing

Slime molds were once classified as fungi, but are now considered to be part of the kingdom Protista. They are eukaryotic organisms that are neither plants, animals, nor fungi. Slime molds have a complex life cycle that includes both single-cell and multicellular stages. They are commonly found in moist, shaded areas like forest floors feeding on microorganisms and dead plant material.

And, despite lacking a brain or nervous system, slime molds can solve the Traveling Salesman Problem, one of the most notorious optimization challenges in computer science.

The Traveling Salesperson Problem (TSP) involves finding the shortest possible route that visits each city in a salesperson’s territory exactly once and returns to the origin city. As the number of cities increases, the problem becomes exponentially more complex, making it difficult for traditional computers to solve efficiently.

In a remarkable experiment published in Science in 2010, researchers from Hokkaido University in Japan demonstrated that the slime mold Physarum polycephalum could approximate solutions to this problem:

1. They placed oat flakes (which slime molds love to eat) on a gel in positions representing cities in Japan.
2. They introduced the slime mold to the "map" and allowed it to grow.
3. The slime mold created an efficient network connecting the food sources, closely resembling the actual rail system connecting those Japanese cities.

The slime mold's solution was comparable in efficiency to computer algorithms specifically designed to solve the TSP.

Slime mold has also shown other problem-solving capabilities including the ability to solve mazes (when placed in a maze with food sources at the entrance and exit, slime mold can find the shortest path between the two points) and to form memories (slime mold can "remember" and anticipate periodic events, such as regular cold shocks, by slowing down its growth just before the expected time of the shock).

These abilities stem from slime mold's decentralized information processing. As it explores its environment, it leaves behind a trail of slime. This trail acts as a form of external memory, allowing different parts of the organism to communicate and make collective decisions.

This brain-free natural information processing system has connections to a number of areas of computer science:

1. Parallel computation: Slime molds process information in a highly parallel manner. Different parts of the organism can simultaneously explore various paths or solutions, making them efficient at solving certain types of problems.
2. Optimization algorithms: The way slime molds grow and adapt to find optimal paths between food sources has inspired new optimization algorithms. As noted above, these "Physarum algorithms" have been applied to various problems including network design and transportation planning.
3. Multi-objective optimization: In finding paths between multiple food sources, slime molds effectively solve multi-objective optimization problems, balancing factors like path length, reliability, and cost.
4. Memristive properties: Some researchers have found that slime mold protoplasmic tubes exhibit memristive properties, suggesting potential applications in bioelectronics and neuromorphic computing.

Research into slime molds and computation is part of a broader field called natural computing or unconventional computing, which explores novel substrates and methods for information processing beyond traditional silicon-based computers.

For example, DNA computing uses the principles of DNA replication and molecular biology to perform calculations. In DNA computing, information is encoded in DNA sequences and computations are carried out through biochemical reactions. This approach has several potential advantages over traditional computing including:

1. Massive parallelism: DNA molecules can perform many calculations simultaneously, potentially solving certain problems much faster than traditional computers.
2. Energy efficiency: DNA computations require very little energy compared to silicon-based computers.
3. Information density: DNA can store an enormous amount of information in a tiny space, far exceeding current electronic storage capabilities.

One notable result in this field was published by Leonard Adleman in Science in 1994, where he used DNA molecules to solve an instance of the Hamiltonian path problem (a variation of the TSP). Adleman’s work demonstrated the potential of DNA computing for tackling complex mathematical problems.

As demonstrated by slime mold and DNA computing, studying unconventional computing is important for a variety of reasons:

1. Pushing boundaries of computation: Unconventional computing challenges our fundamental understanding of what computation is and how it can be achieved. This could lead to entirely new paradigms in computer science and information processing.
2. Overcoming limitations of traditional computing: As we approach the physical limits of silicon-based computing, unconventional methods may offer ways to continue improving computational power and efficiency.
3. Energy efficiency: Many unconventional computing systems, like biological computers, are extremely energy-efficient compared to traditional electronic systems. This could lead to more sustainable computing solutions.
4. Solving complex problems: Some unconventional computing methods, such as quantum computing or DNA computing, have the potential to solve certain complex problems much faster than classical computers.
5. Inspiration for new algorithms: Natural computing systems often inspire new algorithms that can be implemented on traditional computers, improving our problem-solving capabilities.
6. Interdisciplinary research: Unconventional computing bridges multiple fields including computer science, biology, physics, and chemistry, fostering interdisciplinary collaboration and innovation.
7. Adaptability and resilience: Many unconventional computing systems, especially those inspired by biological processes, are highly adaptable and resilient, which could lead to more robust computing systems.
8. Miniaturization: Some forms of unconventional computing, like molecular computing, operate at extremely small scales, potentially enabling new applications in nanotechnology and medicine.
9. Understanding natural information processing: Studying unconventional computing helps us better understand how information processing occurs in nature, from single cells to complex ecosystems.

By exploring these alternative approaches to computation, we not only expand our technological capabilities but also deepen our understanding of the fundamental nature of information and computation itself.

And, as an aside, in case you are tired of your MacBook for some reason, a British computer scientist named Andrew Adamatzky has written a user’s manual for running a slime mold computer.

– Geoffrey W. Smith



First Five
First Five is our curated list of articles, studies, and publications for the month.
‍
1/ A Genealogy of Technology and Power Since 1500
Calculating Empires is a mind-blowing, large-scale research visualization exploring how technical and social structures co-evolved over five centuries.

2/ Techno-optimism
As the VC world has suddenly waded into presidential politics with each side claiming superior support for the tech space, this piece on techno-optimism by Noah Smith is an interesting reminder why getting tech policy right may matter.

3/ Less techno-optimism
This piece in the MIT Technology Review discusses three books that tackle the topic of technological complexity and the wicked problems it creates.

4/ A new financing tool
The D-SAFE (Development Simple Agreement for Future Equity) is modeled after Y Combinator’s SAFE instrument but designed specifically for companies looking to finance development and growth.

5/ Ghosts in the machine?
A new essay in the journal Philosophical Studies considers whether it would be desirable for AI to develop consciousness. "My aim is to contribute to two goals: Firstly, to reduce the risk of inadvertently creating artificial consciousness; this is a desirable outcome, as it's currently unclear under what conditions the creation of artificial consciousness is morally permissible. Secondly, this approach should help rule out deception by ostensibly conscious AI systems that only appear to be conscious.”



Did You Know?
In this section of our newsletter, we hope to demystify common terms and notions in our work as investors.

Term Sheet
A venture capital term sheet is like the first draft of a business deal between a startup and an investor. It's the blueprint for the investment, laying out important details such as how much the startup is worth, how much money the investor is putting in, and the rules of engagement. Think of it as a handshake agreement that sets the stage for the more detailed legal documents to follow. It's a crucial step in the process, helping both parties get on the same page before sealing the deal.

A term sheet typically details like valuation, investment amount, liquidation preferences, governance, antidilution clause,  and other key terms.

– Haiming Chen  & Dylan Henderson