Biology & AI
Artificial intelligence that learns from living beings.
We translate how living beings perceive, adapt and learn into methods for technology. This page explains how that works and why it makes a difference.
Basics
Intelligence does not only sit in the head.
A living being does not think separately from its body. Eyes, skin and muscles do not just deliver data. They take a large part of the work off the brain. This very interplay of senses, body and learning is what we call embodied intelligence.
Preprocessing in the sensor
The retina does not pass on every pixel detail. It filters and condenses before the brain receives anything at all.
Reaction without a detour
Reflexes and fast feedback react before a long deliberation is needed. This saves time and energy.
The body as part of the solution
Elastic tissues balance out disturbances by themselves. The body solves part of the task without any calculation.
Try it out
The eye does not report everything. Only what matters.
A camera sends every pixel in every cycle. The eye works differently: already in the retina, nerve cells respond to edges, contrast and motion. Below, a single dot moves. See how many reports each system needs for it.
Classic camera
Reports every point in every cycle.
Eye, event-based
Reports only what changes.
Less data
Calm parts of the image produce no reports. Only change counts.
Faster reaction
Whoever has to evaluate less can react sooner.
Less energy
What is not calculated needs no electricity.
The eye has more than 120 million light-sensing cells but only about one million nerve fibers to the brain. The condensing begins already in the retina. The illustration above is greatly simplified.
Why it matters
More computing power does not solve every problem.
Many technical systems do not hit the limit of computing power but the limit of their design. Three typical causes:
Data flood
Classic sensing constantly delivers complete images, even when nothing changes. This costs time, energy and computing effort.
Rigid procedures
Fixed programs work as long as the environment fits the plan. As soon as it deviates, the reserves are missing.
Separate development
When sensing, body and software are created separately, they do not fit together optimally and the advantages are lost.
We develop them together, modeled on biology.
Four capabilities
What we take from living beings.
Pick a capability. You see how it works in nature and what we make of it.
Perception
Eyes and sensory cells mainly report changes and give the brain a condensed picture of the surroundings.
Event-based and spatial perception: little data, fast evaluation, depth in real time.
Adaptation
Tissue, posture and movement adjust to ground, load and speed, often without conscious control.
Compliant components and controls that balance deviations through interplay instead of counter-steering rigidly.
Learning
Experience changes behavior: movements become more precise and economical with practice.
Embodied AI that learns from experience with its own body and keeps developing.
Reaction
Fast feedback reacts before a long deliberation is needed, even with incomplete information.
Decisions in real time that are based on perception and stay viable even under uncertainty.
Perception in detail
Seeing like a living being.
Perception is the starting point for everything else. We work on three building blocks.
Building block 1
Detecting depth
Two eyes see an object from slightly different angles. The difference reveals how far away it is. We build methods that derive distances in real time from two viewpoints.
- Two viewpoints produce a spatial picture
- Distances are derived without extra aids
- The evaluation runs in real time
Building block 2
Capturing change
The retina mainly reports what changes. Event-based perception follows this principle: a measuring point reports only when something changes, not in every cycle.
- In a calm scene hardly any data arises
- Movement is detected very quickly
- The computing effort drops
Building block 3
Understanding context
Single measurements are not yet understanding. Only when depth, motion and touch come together into a whole picture can a system judge what is happening.
- Measurements are linked with each other
- What matters is separated from what does not
- The whole picture is the basis for learning
Seeing in the dark
Why colors disappear at dusk.
The eye works with two kinds of light-sensing cells. Depending on brightness, one or the other takes over. In between lies a transition in which seeing changes noticeably.
Day
The cones take over. We see colors and fine detail.
Twilight
Cones and rods work together. Colors fade, contrasts gain importance. Red looks darker, blue brighter.
Night
The rods take over. They are very sensitive to light but see only light and dark.
Technology that works reliably only in good light reaches its limits exactly in this in-between range. The eye adjusts. We study how this adjusting can be transferred to perception technology.
Simplified illustration. Full dark adaptation takes about half an hour.
Evolution as an engineer
Evolution did not find one solution. It found many.
Eyes arose several times independently in the history of life. Each solution suits a habitat and a task. That is why we learn from the diversity and not from a single model.
Compound eye
Insects assemble a picture from many small single eyes. In return they detect motion and flicker far more finely than humans.
Raptor eye
Birds of prey spot prey from a great distance. Their eye is built for the highest sharpness.
Deep-sea eye
Where hardly any light arrives, sensitivity counts. Such eyes still extract information from tiny amounts of light.
And humans? They get a bit of everything: colors by day, contrast at dusk, motion at the edge of the visual field and a brain that fills in gaps.
Cycle
Perceive, process, act, learn.
Embodied AI is not a one-way street. Every action changes what is perceived next. Learning arises from this cycle.
- Perceive: Senses capture the surroundings and the state of the body.
- Process: Signals are condensed and classified.
- Act: The body reacts, elastically and adapted.
- Learn: Experience improves the next pass.
Limits and opportunities
Where technology stumbles today, our work begins.
Four weak points hold back today's systems. In each, nature shows a different answer.
Energy
- Today
- Data centers and high-performance computers need a great deal of electricity for perception.
- Nature as a model
- The brain gets by with about 20 watts.
- Our approach
- Processing that becomes active only when something changes.
Robustness
- Today
- Fog, glare, dust or tiny disturbances throw systems off track.
- Nature as a model
- The eye has blind spots, the brain balances them out.
- Our approach
- Systems that reckon with uncertainty instead of demanding perfection.
Context
- Today
- An image is at first only a set of points without meaning.
- Nature as a model
- Living beings understand what they see because they have experience with the world.
- Our approach
- Linking perception, movement and touch into a whole picture.
Learning
- Today
- Many methods need huge amounts of data to learn something simple.
- Nature as a model
- Living beings learn from few experiences and in context.
- Our approach
- Embodied AI that learns from experience.
Approach
How observation becomes technology.
Every principle goes through five steps with us. Only when one step is backed up does the next follow.
Observe
We study how living beings solve a task, for example how they estimate distances or sense touch, at the level of cells, tissues and behavior.
Involved: biology, neuroscienceModel
From the observations come models that can be verified. They describe what is essential for the function and what is not.
Involved: biology, biomechanics, AITranslate
The model is transferred into technology: into sensing, components and algorithms that use the same principle without copying the living being.
Involved: engineering, AITest
The results are checked independently. In 2026, tests at leading universities take place for this.
Involved: all disciplines, external partnersImprove
Test results flow back into the models. Step by step this creates a system based on proven findings.
Involved: all disciplinesNumbers of seeing
What happens in the eye and in the head every second.
Rounded reference values from the scientific literature. Individual values vary by source and person.
Terms
Briefly explained.
Bioinspired
Functional principles from biology are understood and implemented technically. We do not copy living beings.
Embodied AI
AI that is connected to a body and senses and learns through experience with them.
Event-based perception
Sensors report only changes instead of constant full images.
Stereo depth
Distance from two viewpoints, as when seeing with two eyes.
Neuromorphic
Technology whose design is modeled on nervous systems, for example on the processing of single impulses.
Real time
Processing so fast that the reaction fits the current situation.
Retina
The layer in the eye where light is converted into nerve signals. Selection already begins here.
Cones and rods
Cones see color in light, rods see light and dark in low light.
Dark adaptation
The gradual adjustment of the eye to low light. It takes about half an hour.
Mesopic vision
The transition range between day and night vision, for example at dusk.
Parallel processing
Many signals are processed at the same time instead of one after another. This is how the brain works.
Robustness
The ability to keep working stably even with disturbances and uncertainty.
Who works on it
Questions of biology and engineering are asked together.
Questions
Frequently asked questions about Biology & AI.
What does “Biology & AI” mean at Eyroq?
We understand how living beings perceive, adapt and learn, and transfer these principles into AI methods and sensing. Biologists and engineers work together from the very start.
What is embodied AI?
Embodied AI is connected to a body and senses. It learns from the experience it gains with them, and not only from abstract data.
Why does everything begin with perception?
What a system cannot capture, it can neither understand nor respond to. Good perception delivers little but meaningful data and is thus the basis for learning and reaction.
What is event-based perception?
Instead of constantly delivering full images, the measuring points report something only when something changes. This matches the principle of the retina and saves data and time.
How does this differ from classic AI?
Classic AI is often developed separately from the body and trained on large amounts of data. We develop perception, body and learning together so that they fit each other.
Which disciplines work on Biology & AI?
Biology, neuroscience, biomechanics, engineering and artificial intelligence. Questions about perception and learning arise at the interfaces of these fields, which is why they work together here.
How do you check whether a principle is really transferable?
We first describe the principle as a model, then test it on components and finally have it cross-checked externally. The independent tests at leading universities are planned for 2026.
Can I talk to you about Biology & AI?
Gladly. Write to us at info@eyroq.com and briefly describe what interests you. We will get back to you.
Why is the eye a model for technology?
Because it filters the essentials out of huge amounts of data very efficiently. Already in the retina, edges, contrast and motion are emphasized, and only a small part of the information goes to the brain.
How much energy does the brain need?
About 20 watts, roughly as much as a small light bulb. For comparable tasks, technical systems often need a multiple of that.
What happens when seeing at dusk?
Cones and rods work together. Colors fade, contrasts become more important, and after about half an hour the eye is fully adapted to the dark.
What are neuromorphic chips?
These are chips whose design is modeled on nervous systems. They process signals as single impulses and need energy mainly when something actually happens. They are an active field of research.
Does bioinspired mean you copy nature?
No. We understand the principle and implement it with technical means. It is about the how, not about rebuilding a living being.
Keep exploring.
Questions about Biology & AI? Let’s talk.
Tell us briefly what it is about. We will get back to you.