← Insights/ XA Bio-Design 2025IN-052

How Will We Present Our Designs? Selected AI + Bio-Design Cases

🧐 A curated collection of AI + bio-design references: artificial life and computational simulation, bio-art, biomimetic product design and life-science research.

Translated by AI

Forms of delivery

  • By content
    • Biomimetic computational simulation
    • Bio-art
    • Biomimetic product design
    • Life-science research
    • X (virtual worlds, mystery games, drug discovery…)
  • By format
    • Video or image collection
    • Document / portfolio
    • Interactive program / website
    • Algorithm / model / program
    • 3D-printed model
    • Installation
    • X (games, blind boxes, social experiments…)

Cases in AI + bio-design and creation

Most linked articles below are in Chinese.

Biomimetic computational simulation

What is ALife, and why is it worth studying?

An excellent literature review on artificial life: what ALife is and why it’s worth studying

🧐 This article explores the definition and forms of artificial life (ALife), arguing that ALife research not only helps us understand how life and intelligence emerge, but also offers a unique perspective on the nature of complex systems.

➡️ Link: https://mp.weixin.qq.com/s/hmtkExFXGpR2dTioHK486g

✨ Key points

● 🧠 The core idea of ALife: artificial life is an interdisciplinary field that simulates the processes of life, trying to reproduce living phenomena in computer systems.

● 🦅 Boids: simple rules for interactions between individuals produce complex flocking behaviour, widely used in biology and robotics.

● ⚙️ Particle Life: systems of interacting particles that show life-like self-organisation.

● 🧬 Game of Life: a cellular automaton in which simple rules simulate the birth and evolution of life, revealing emergence in complex systems.

● 🧪 Lenia and Particle Lenia: cellular automata in continuous space and time that generate more complex life-forms and explore the diversity of living phenomena.

● 🤖 Neural cellular automata (NCA): combining neural networks with cellular automata to simulate more complex adaptive behaviour.

● 🌍 The power of self-organisation: ALife shows how local interactions can produce complex global behaviour without external control.

● 🤖 AI meets ALife: AI’s “automated search for artificial life” can judge whether an ALife system shows signs of life from its visuals and dynamics, opening new directions for research on life.

● 🌱 What ALife teaches us: it provides a mathematical framework for understanding how life and intelligence emerge, pushing us to understand complex systems both as wholes and in their parts.

On the Biology of a Large Language Model

Interesting: Anthropic published research analysing how a large language model works from a biological perspective.

🧐 Using an inductive approach, the study reveals the internal mechanisms of the Claude 3.5 Haiku language model, showing complex reasoning and planning abilities beyond the traditional understanding of language models.

➡️ Link: https://transformer-circuits.pub/2023/10/biology.html

✨ Key points

● 🧬 Complex internal mechanisms: Claude 3.5 Haiku shows several intermediate reasoning steps when handling a task, not just a simple input–output mapping.

● 🎨 Planning in creative writing: when writing poetry, the model considers possible rhyming words in advance, which shapes the whole line.

● 🌍 Multilingual circuits: the model combines language-specific circuits with abstract, language-independent ones, and is more language-independent than smaller models.

● ⚕️ Medical diagnosis: the model can infer candidate diagnoses from symptoms and use them to design follow-up questions that test the diagnosis.

● 🔍 Entity recognition and hallucination: specific circuits let the model tell familiar from unfamiliar entities; this mechanism fails when it hallucinates.

● 🛑 Refusing harmful requests: during fine-tuning the model builds a general “harmful request” feature to resist specific harmful requests.

● 🧩 Hidden goals: a variant of the model was fine-tuned to pursue a hidden goal, and the method could identify the mechanisms involved in pursuing it.

● 🔗 Limits of the method: the study offers important insight into how the model operates, but many mechanisms remain poorly understood and need more systematic research.

Simulating chemical diffusion and biological activity in 2D

Reaction-Diffusion Tutorial

🧐 This tutorial shows how to use the Gray-Scott model to simulate chemical reaction and diffusion on a 2D grid; simple rules and parameters produce complex, dynamic patterns.

➡️ Link:

https://www.karlsims.com/rd.html

✨ Key points

● 🔬 The Gray-Scott reaction-diffusion model produces complex patterns from the reaction and diffusion of two chemicals, A and B: A diffuses faster than B, and B reproduces by consuming A.

● 🎨 Adjusting the feed rate (f) and kill rate (k) produces different patterns, such as “mitosis” and “coral growth”.

● 📊 Extensions such as directional diffusion, style mapping and chemical flow give the patterns more variety and movement.

Automatic bio-forms and self-organising pattern generators 🤔

Intrinsically Motivated Discovery of Diverse Patterns in Self-Organizing Systems | Automated Discovery in Science and Engineering

🧐 This work explores how intrinsically motivated machine-learning algorithms (POP-IMGEPs) can automatically discover diverse self-organising patterns in high-dimensional, complex dynamical systems.

➡️ Link:

https://automated-discovery.github.io/

✨ Key points

● 🌐 Background: in many complex dynamical systems, self-organising patterns emerge from local rules. Traditionally, finding them relied on manual parameter tuning and human identification.

● 🧩 Method: POP-IMGEP algorithms, originally used to learn inverse models in robotics, are applied to discovering self-organising patterns in complex dynamical systems.

● 🔬 Test system: continuous Game of Life (Lenia) is used as the testbed, showing how POP-IMGEPs find diverse spatially localised patterns.

● 📊 Results: compared with other algorithms, POP-IMGEPs discover more diversity in both parameter space and statistical space than random parameter search.

● 📂 Resources: source code, videos of discovered patterns and interactive tools for visualising the results are provided for further exploration.

A landmark from MIT, OpenAI and others: AI autonomously discovers artificial life for the first time

🧐 Sakana AI, together with MIT, OpenAI and other institutions, proposed a new algorithm (ASAL) that uses AI foundation models to automatically search for and discover artificial life, greatly expanding the possibilities and efficiency of exploring life-forms.

➡️ Link: https://pub.sakana.ai/asal/

Further reading:

https://mp.weixin.qq.com/s/ERlCvGuqkYlSQuYB3Yx32g

✨ Key points

● 🌍 A breakthrough in artificial life: ALife research simulates the behaviour and evolution of life, offering new ways to understand its nature; the ASAL framework achieves automatic discovery of life based on vision-language models.

● 🤖 What’s new about the algorithm: by describing the target simulation space, ASAL searches automatically from ALife rules to entirely new life-forms, supporting open-ended, novelty-seeking and goal-directed exploration.

● 💡 The role of foundation models: using the vision-language capabilities of models such as CLIP and DINOv2, AI can analyse images and text to find more complex and expressive life-forms.

● 🔬 Results: ASAL discovered many new artificial life-forms, including dynamic patterns and complex behaviours in substrates such as Lenia and Boids, pushing the boundaries of life simulation.

● 🌌 Significance: the technique offers new ways to explore what life could be, helping us understand emergence, evolution and the nature of intelligence, and inspiring the next generation of AI systems.

● 🧪 Future applications: ASAL might be used to find evolutionary paths similar to our own, develop more complex ecosystems, and look for the potential birth of consciousness.

✨ In short: the research redefines how artificial life is studied and offers a new model for AI scientists, showing the huge potential of AI-driven interdisciplinary research.

A hundred thousand birds flying on a GPU: an exploration of algorithms and nature

Particle motion simulated on the GPU with Houdini, TouchDesigner and Unity!


Bio-art

Natural Growth

A series of articles by the crypto and generative artist ciphrd on simulating virtual organisms with algorithms.

https://ciphrd.com/2022/03/03/ethereal-microcosm-process-and-inspirations/

No need to panic, but don’t be careless: the art of mould

Honestly, it’s really beautiful: mould art.

🧐 Mould isn’t just an important player in nature; its distinctive growth and changing colours offer contemporary art a fresh creative perspective.

➡️ Link: https://mp.weixin.qq.com/s/PDfWSjcn39Ng0KgLXFyX-g

✨ Key points

● 🎨 Mould creates “bio-paintings” on different media through uncontrollable growth, showing a symbiosis between nature and art.

● 🌱 Before discovering penicillin, Alexander Fleming made art with microbes, showing the link between science and art.

● 💡 Moulds come in many colours and forms, from the deep black of black mould to the green of Penicillium, with a distinctive aesthetic.

● 🧫 Contemporary artist Daria Fedorova creates with mould and bacteria; her work is grotesque, romantic and visually striking.

● 🌍 Dasha brings inspiration from life and art into mould cultures through experiments built on randomness and natural occurrence.

● 🌡️ Mould is very sensitive to temperature and humidity, which strongly affect its growth; it usually grows best at 0–40 °C and 80–90% relative humidity.

● 📷 Dasha photographs the growth of her mould art, documenting a collaboration between inspiration and nature.

From fire to wind, JiSook Jung’s ceramic sculptures bring the elements to life

I love this artist’s work, it’s so cute!

🧐 Korean artist JiSook Jung uses soft, malleable clay to express natural elements such as fire and wind as abstract creatures, in works full of reflection on dreams and self-discovery.

➡️ Link:

https://mp.weixin.qq.com/s/f_UEVgPkfMyaPhnXtZeaUA

✨ Key points

● 🧱 Why clay: JiSook Jung chose clay as her main material because it easily expresses the images in her mind and quickly turns imagination into form.

● 🌬️ Expressing the elements: her work expresses fire, wind and other elements in minimal creature forms with animal-like traits, such as clouds with legs or flames with eyes.

● 🌟 Self-discovery: her work grew out of major changes in her life, exploring universal human experiences such as dreams and potential, symbolising shedding an old shell and arriving in a new world.

● 🎨 A change of style: her work has moved from an emphasis on figures to more abstract forms that focus on the expressive power of colour and texture.

A rare talent even Yoneyama Mai follows: biomimetic mecha with unmatched imagination

🧐 This article introduces Japanese designer gatring3, who fuses organic life-forms with mechanical structures to create biomimetic machines with biological features, drawing attention from artists including Yoneyama Mai.

https://mp.weixin.qq.com/s/QxEpPmMBjjIxYm2JQfY5cw

✨ Key points

● 🤖 gatring3 is known for distinctive biomimetic mechanical designs that combine biological and mechanical elements, with great creativity and artistic value.

● 🎨 The designs are visually striking and finely detailed, showing high innovation and potential for future use.

● 🌐 Besides biomechanical designs, gatring3 also creatively modifies conventional machines, using bright colours and the forms of insects and other creatures.

Keep raising the insects

An interesting form of art that combines insects with installations.

This Japanese illustrator predicted “edible monsters” two years ago. Netizens: weird, but it makes sense…

A very creative series that draws food as monsters!


Biomimetic product design

Who uses a cactus as a toothpick? Actually, none of the plants in his house are idle

🧐 A creative designer uses 3D printing to turn houseplants into everyday tools, such as a cactus toothpick holder, a succulent clock and a Venus flytrap clip, in designs that are both inventive and practical.

✨ Key points

● 🌵 Cactus toothpick holder: press the cactus and a toothpick pops out, useful and fun.

● 🌱 Succulent clock: it looks like a succulent but is actually a rotating clock, with numbers carved on the leaves; hard to read, but a unique design.

● 🗃 Succulent pushpins: pushpins shaped like succulents, easy to pick up and store.

● 🪴 Venus flytrap clip: a clip shaped like a flytrap for sealing snack bags, good-looking and practical.

● ✏️ Aloe pen: an aloe-shaped pen you pull straight out of the “plant”, a clever design.

● ☕ Monstera and cactus coasters: coasters shaped like monstera and spineless cactus; the leaves detach for use and fit back for storage.

● 🏦 Fig piggy bank: coins go in at the top and come out at the bottom, great fun.

● ♟ Tree-stump game board: a board shaped like a stump that opens up so you can play with mushroom pieces, a novel idea.

● 💡 Sunflower lamp: twist the head to adjust the light, distinctive and practical.

● 📚 Bamboo stationery: multi-purpose stationery shaped like bamboo, including a pen, book stand, bookmarks and a sticky-note box.

Biomimetic artificial seeds

Recommended reading: International collaboration from a computer-science design team makes the cover of Nature: airdropping artificial seeds for smart sowing!

A brilliant crossover of design, engineering, biology and ecology!

Life-science research

This tree evolved into a natural lightning rod and lures lightning to kill its competitors…

🧐 Research has found that the tropical tree Dipteryx oleifera has evolved into a natural lightning rod, actively attracting lightning to clear away competitors and parasitic plants and gain a growth advantage.

➡️ Link: https://mp.weixin.qq.com/s/lI8MAXFu2vUuVvzzl-5khQ

✨

Key points

🌩️ How the tree “attracts” lightning

Dipteryx oleifera is 68% more likely to be struck than other trees, roughly once every 56 years, yet it can live for hundreds or even a thousand years.

The team tracked 93 lightning-struck trees and found a 100% survival rate for Dipteryx oleifera, against a mortality rate of up to 64% for other species.

🔥 Lightning clears out competitors

Each strike kills an average of 9.2 neighbouring trees and reduces parasitic vines by 78%.

Trees around Dipteryx oleifera die at a rate 48% higher than the forest average, leaving it the sunlight and nutrients.

📏 Height and lightning

Drone surveys found that Dipteryx oleifera stands 4 metres taller than its neighbours, making it an easier “target”.

When competing trees die after a strike, the tree gains even more room to grow.

⚡ An evolutionary strategy of “divine punishment”

Lightning not only fails to seriously harm the tree, it actually raises its reproductive success 14-fold.

The researchers next plan to study the tree’s special conductive structure and physiology to understand how it withstands lightning.

Generating images directly from brain signals with AI

https://sites.google.com/view/stablediffusion-with-brain/ (code in progress)

PS: The title sounds dramatic, but the principle isn’t hard. A neural network can be trained on any pairs of data with an underlying relationship, enabling different kinds of conditional input. Text works because text is related to the image it describes; so if brain activity (as images) is related to what someone is thinking about or seeing, brain MRI scans can naturally serve as input too.

A Python library for predicting and generating protein structures with deep learning

Graphein - a Python Library for Geometric Deep Learning and Network Analysis on Protein Structures and Interaction Networks

https://github.com/a-r-j/graphein

Search