#neuromorphic — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #neuromorphic, aggregated by home.social.
-
New paper out 🧠
1. We showed that an #EEG #neuromorphic classifier achieves comparable GPU results using 1/3 of electricity
2. 1/f slope (E/I inbalance) was among top features, and we showed the value of it also with #Alzheimer neurosimulations with #Nest
https://www.sciencedirect.com/science/article/pii/S0925231226022885 -
New preprint on #neuromodulation in #SpikingNeuralNetworks of interest to people in #neuroscience and #neuromorphic computing.
Headline result: neuromodulation gives a huge boost to performance, particularly in noisy environments, and reduces energy consumption by reducing neuron and spike counts.
We also have tutorials and codes to integrate easily into your models.
-
Sound waves could speed up #neuromorphic computing, according to new research. A new acoustic synapse can have multiple parallel computations, with lower power costs than conventional electronics.
spectrum.ieee.org/neuromorphic... -
New #neuromorphic #SpikingNeuralNetworks paper with Pengfei Sun, Zhe Su, @achterbrain, @giacomoi and Danyal Akarca:
https://neural-reckoning.org/pub_dual_memory_pathways.html
Long story short, we found a neat new trick for augmenting spiking neural networks with a tiny separate memory buffer, that gives a huge boost to performance on tasks with temporal structure at a very low memory cost (state-of-the-art performance with 40-60% fewer parameters). It's also very neuromorphic hardware-friendly and gives us 4x increased throughput and 5x improvement in energy efficiency compared to state-of-the-art.
We think this is (a) a great example of where an abstraction inspired by biology can lead to networks that are both algorithmically effective and hardware-efficient, (b) shows the power of doing hardware-software co-design for neuromorphic computation.
-
RE: https://mathstodon.xyz/@gconstantinides/116448779816230899
We're hiring for a project on #SpikingNeuralNetworks and #neuromorphic computing, to start in October this year, for 36 months. Can hire at pre- or post-PhD level. Feel free to email me informally, or apply at the link below. Please do share with your networks if you know someone who would be interested.
#ComputationalNeuroscience -
#Zoomposium with #ThomasMetzinger Part 2: 🤖 #Self-Models, #neuromorphic #AI & the Future of Our #CultureofConsciousness 🧠
What is the “self”—and could a #machine one day develop its own #consciousness?
In the second part of the discussion, the #SelfModelTheory of #Subjectivity takes center stage.
🎥 https://youtu.be/Ibf2mrZ3uzM
📎https://philosophies.de/index.php/2026/03/15/selbstmodelle-metzinger/
#ArtificialIntelligence #PhilosophyOfMind #Neuroethics #KIEthics #TechnologyAndSociety #FutureOfAI #Philosophy
-
9/
The Constraints: Physics, chemistry, energy consumption.The Emergent Architecture: The Octopus.
The paper is essentially trying to turn the "Jazz Band" chaos of evolution into a repeatable engineering formula. It’s "old wine in a new bottle," but the bottle is now high-performance computing.
#AI
#neuromorphic
#DistributiveIntelligence
#evolution
#octopus -
8/
3. The "Neuromorphic" Hardware AngleThe "new" part is often the Hardware (Neuromorphic). We are finally building chips that can actually handle these distributed rules without a central "Boss" CPU bottlenecking everything.
The "Mind Club" Connection
If you feel like you’ve heard this before, it’s because you have—it’s the story of Evolution. -
7/
2. Declarative vs. ProceduralMost AI today is "Procedural"—it’s fed millions of examples to learn a procedure. A "Declarative Constraint" system is different. It’s like telling a robot: "You are not allowed to touch the walls," and letting it figure out navigation, rather than showing it 10,000 videos of robots not touching walls.
-
6/
1. From "Bio-Inspiration" to "Mechanical Necessity"Usually, AI researchers try to mimic the brain (Neural Networks). Kinney is suggesting we stop trying to copy the brain's look and instead copy its limitations. He’s arguing that if you set the right "Universal Constraints," the system is forced to develop a brain-like structure because that's the only mathematically efficient way to solve the problem.
-
5/
The core idea of emergence from rules is nothing new. It’s what John Conway was doing with the Game of Life in 1970, and it’s how Stephen Wolfram has viewed the universe for decades.What makes this specific paper by Kinney "new" (or at least a fresh take) isn't the concept of emergence itself, but the application:
-
4/
The Big Idea
Stephen Kinney is essentially arguing that we can create "Artificial Life" (or at least more flexible AI) by focusing on Limitations (Constraints) rather than Commands. -
#Neuromorphic explained:
In plain language, this paper is describing a way to build a digital "brain" (a neuromorphic architecture) not by programming every step, but by simply giving the system a set of Universal Rules (constraints) to follow.
Think of it like this:
The "Lego" Analogy
Traditional AI: You give the computer a massive instruction manual on how to build a specific castle, brick by brick. If you want a tower, you have to code "tower." -
Ok. I think the last widget I'll be building for my Bio-Inspired #AI and Optimization course this semester.
Spiking Neural Network ( #SNN ) explorer
LIF Neuron model, Rate coding, STDP, Survey of HW&SW implementations, #Neuromorphic applications & learning
https://tpavlic.github.io/asu-bioinspired-ai-and-optimization/spiking_neural_networks/snn_explorer.html -
#Zoomposium with #ThomasMetzinger Part 2: 🤖 #Self-Models, #neuromorphic #AI & the Future of Our #CultureofConsciousness 🧠
What is the “self”—and could a #machine one day develop its own #consciousness?
In the second part of the discussion, the #SelfModelTheory of #Subjectivity takes center stage.
🎥 https://youtu.be/Ibf2mrZ3uzM
📎https://philosophies.de/index.php/2026/03/15/selbstmodelle-metzinger/
#ArtificialIntelligence #PhilosophyOfMind #Neuroethics #KIEthics #TechnologyAndSociety #FutureOfAI #Philosophy
-
Together with Clara Wanjura, we are delighted to announce the second instalment of our workshop on neuromorphic computing taking place 10-12 June at the @MPI_ScienceOfLight Max Planck Institute for the Science of Light in Erlangen, Germany.
👉 Registration and abstract submission will be open until 12 April.
For further information, please visit the event website: https://indico.mpl.mpg.de/event/27/
-
Between commodification, edge computing, and (eventually) #neuromorphic approaches (instead of oodles of multiply-accumulate #hardware), some of the #AI spend is going to look rather foolish eventually
https://asymco.com/2026/03/10/the-most-brilliant-move-in-corporate-history/ -
Decima-8: Нейроморфная архитектура, оперирующая уровнями энергии
Современные нейроморфные системы сталкиваются с двумя независимыми проблемами. Проблема 1: Кодирование информации Бинарные спайковые сети (SNN) передают градации сигнала через: Частотное кодирование (множество тактов на одно значение) Увеличение количества линий передачи Проблема 2: Аппаратная реализация Аналоговые мемристорные кроссбары обещают естественную нейроморфность, но содержат следующие проблемы: Шум и дрейф параметров Недетерминизм вычислений Каждый чип требует индивидуальной калибровки Традиционные Network-on-Chip (NoC) добавляют overhead: ~40% площади кристалла уходит на маршрутизаторы ~70% энергии тратится на пересылку данных, а не вычисления Decima-8 предлагает: Level16: кодирование уровня активации (0..15) в одном такте на одной линии. Это компромисс между бинарным представлением и аналоговой непрерывностью. Цифровые кроссбары (эмуляция мемристорных матриц): детерминизм, воспроизводимость, отсутствие шума Эстафетную активацию вместо пакетной маршрутизации: тайлы не передают данные друг другу, активация распространяется через граф зависимостей Результат: фиксированная задержка, предсказуемое поведение, 0% площади на роутеры.
-
Neuromorphic computing has proved to be good at complex mathematics, against expectations. New #neuromorphic hardware from #Intel can solve differential equations using the finite element method. spectrum.ieee.org/neuromorphic...
-
# 🌟 **φ⁴³ AQARION-BUNDLE TikTok Presentation** 🎬 (3773 chars)
```
🚀 60s → ENTERPRISE RAG DOMINATION! 🔥PROBLEM:
Enterprise RAG = $900K/YR 😱
77% accuracy 📉 3.2s latency 🐌SOLUTION: φ⁴³ AQARION-BUNDLE! 💎
94.1% accuracy 📈 0.9ms latency ⚡
$85/MO vs $900K/YR → $450K SAVINGS! 💰ONE COMMAND:
curl | python3 → PRODUCTION LIVE! 🎯73-NODE HYPERGRAPH DASHBOARD 🖥️
Three.js LIVE φ-heatmap! ✨
Green nodes = φ=1.9102 locked ✅
Edge glow = 0.9ms latency 🌟
#aquarius #neuromorphic #hypergraph -
#Neuromorphic computing can now solve partial differential equations - a non-obvious, but exciting use case beyond #neuralnetworks