Portuguese 
English 
Spanish 
4 min of reading 
0 comments 
Artificial neurons made from flexible polymer generate realistic signals in the brain, trigger living cells, and point to more efficient AI and implants.
Artificial neurons printed by engineers at Northwestern University managed to communicate with living brain cells in tests with mouse brain tissue. The flexible and low-cost devices generated electrical signals realistic enough to trigger responses in real neurons, in a step that brings electronics closer to the nervous system.
The study, which will be published on Wednesday, April 15, in the journal Nature Nanotechnology, points to two areas of impact: implants and brain-machine interfaces, such as neuroprosthetics for hearing, vision, and movement, and brain-inspired computing, with the potential to make AI consume less energy.
Why these artificial neurons gained so much attention
The central point is not just to “imitate” the brain. The artificial neurons printed by the Northwestern team produced more complex signaling patterns, with unique peaks, continuous firing, and bursts, coming closer to the way real neurons communicate.
-
Public school students from Amazonas create soap made with coffee grounds: sustainable product achieves over 60% approval and shows commercial potential and positive environmental impact.
-
While the world still relies on polluting and slow energy, the island of Samsø in Denmark has started producing 40% more energy than it consumes, with 11 turbines, projects of up to 40 million euros, and a sustainable revolution built by just 4,000 inhabitants.
-
Xiaomi launches new smart stove with titanium sensor and automatic shutdown that detects an open flame without supervision.
-
The United States has just reached an unprecedented agreement to use military lasers on American soil — the plan is to create a defense dome that takes down drones at the speed of light and costs cents per shot.
This variety matters because, in many previous attempts, the signals were too simplified. And, when the signal is poor, the output often tends to inflate the structure with more devices, which comes at a high cost in energy consumption. Here, the promise is to do more with less.
From rigid silicon to soft materials that “talk” to the brain
The study starts from a direct comparison: computers increase complexity by stacking billions of identical transistors on rigid, two-dimensional silicon chips. In contrast, the brain operates with three-dimensional, flexible, and constantly remodeling networks, with different neurons performing specialized roles.
The team argues that to get closer to this model, the foundation needs to change. The answer was to bet on soft and printable materials, more similar to the biological environment, instead of trying to force the brain to “speak the language” of silicon.
What was printed and how it became an artificial neuron
The artificial neurons were constructed with electronic inks formulated from molybdenum disulfide (MoS₂) nanoflakes, as a semiconductor, and graphene, as an electrical conductor. To deposit the material, the researchers used aerosol jet printing, printing the inks onto flexible polymer substrates.
The result was a flexible device that, in practice, does not rely on a rigid and expensive structure to function. And this opens up space for electronics that are more compatible with the body and simpler to manufacture. But the key insight came from a detail that many people were trying to eliminate.
The “defect” of the polymer that became the most useful feature
In previous work, the stabilizing polymer in inks was seen as a problem because it interfered with current flow. The traditional solution was to burn the polymer after printing the circuit.
The Northwestern approach was different. Instead of removing the polymer, the team partially decomposed it and used this decomposition to create a localized conductive pathway, concentrating the current in a narrow region. This generated a sudden electrical response, similar to that of a neuron, and allowed for the production of richer and more varied signals.
The test with a living brain that put artificial neurons to the test
To verify if the artificial neurons really interacted with biology, the team collaborated with neurobiologist Indira M. Raman. The electrical signals from the devices were applied to mouse cerebellum slices.
The finding was straightforward: the artificial voltage spikes corresponded to important biological characteristics, such as timing and duration of spikes from living neurons, and were able to trigger activity in real neurons, activating neural circuits similarly to natural signals. It’s the difference between “looking like” a brain and actually activating a brain.
What this could mean for implants and brain-machine interfaces
The study points to potential applications in electronics capable of communicating directly with the nervous system. Among the cited examples are brain-machine interfaces and neuroprosthetics, with possibilities in implants related to hearing, vision, and movement.
The logic is simple: if a device can generate signals with a shape and temporal scale compatible with neurons, it can become a more natural bridge between hardware and body. And when the bridge is more natural, the chance of real use increases.
The connection with AI and the giant problem of energy consumption
Mark C. Hersam, who led the study, points out a bottleneck that has emerged: training AI with massive volumes of data increases energy consumption and puts pressure on infrastructure and cooling.
The proposal inspired by the brain appears as a pathway because, according to the study’s explanation, the brain is much more energy-efficient than a digital computer. The bet is that by mimicking how neurons signal, future systems could perform complex operations using less energy than current technologies. And this affects the heart of AI costs.
Why manufacturing also comes into play
In addition to energy efficiency, the work highlights advantages of the process: the manufacturing of the neuron is described as simple and low-cost. And there is an important environmental factor: since the printing is additive, depositing material only where needed, there is a reduction in waste.
In a scenario where the advancement of AI is hindered by energy, heat, and water use for cooling, every gain in efficiency becomes a strong argument. And this is the part that tends to fuel the next discussion.
Do you think artificial neurons will arrive first as an implant in the body or as hardware for more efficient AI?
Seja o primeiro a reagir!