The skepticism from your end is very understandable. But human brain is made of approximately 86 billion of those pretty little boxes (neurons) of electrical impulses, all governed by same fundamental principles. whether, you have 100 pretty boxes, trillion boxes, the rules remain the same. The real magic is to understand those rules and simulate them using lines of code. Thinking, feeling or experiencing are higher cognitive functions resulting from the interactions of those pretty little boxes. And obviously to get to these cognitive functions, you need more boxes. That's totally fair. But even at small scale, if those rules are correct, you will start seeing early signs which will mature into cognitive functions you are referencing. Think of a development of newborn into an adult. So if you get the rules right, it is completely justified calling it an artificial brain. A toy car is essentially also a car, both governed by same laws of motion. The logic and definition of the use of words(spin) is formally defined in the report. I hope this clarifies some of the skepticism.
Few Anticipatory Questions
• “Isn’t this just ML/randomness?” - No training or gradient descent is used. Only neuron-like rules (graded, subthreshold, action potentials). Outputs are logged and timestamped; anyone can verify them.
• “How do you define ‘original thought’?” - An output is “original” if it was never presented as a stimulus during that system’s lifetime, yet emerges autonomously.
• “What about controls?” - We ran multiple experiments with different genetic parameters; each yielded different system behaviors. One run was deliberately configured as a pure input/output machine, confirming that adaptability is essential for higher functions.
• “Independent replication?” – We are open to live demos (reviewers choose inputs) and will provide full raw outputs. Under NDA, reviewers can also set genetic parameters and observe the system’s lifetime behavior.
• “Why 500 Adaptrons?” – Our approach is milestone-driven: we demonstrate emergence at small scales first (memory, anticipation), then scale gradually (20k, multimodal, 1M).
At JN Research, we are exploring a third path between mainstream traditional AI and descriptive neuroscience. Instead of scaling or optimizing trained function approximators, we build Adaptrons; artificial neurons that behave like biological neurons (subthreshold + graded + Action Potential) and autonomously adapt internally and with other Adaptrons in a system. On this substrate, our small artificial brain Primite 1.02 (500 Adaptrons) now shows: • Original thoughts (novel outputs not seen as stimuli). • Memory formation and consolidation (short/intermediate/long-term). • Anticipation: outputs that appear before the corresponding stimulus is presented. We ran 8 independent experiments with different genetic parameters and share detailed counts, timing, and example outputs. This is not ML training; it’s a principles-first cognitive substrate where higher functions emerge from the interaction rules. Furthermore, we also show that higher cognitive functions do not need bigger models or scale to emerge, we can see their early signs if the fundamental framework allows for it. If you are curious (or skeptical), we have included the full technical report and a data repo with outputs for verification, plus our prior 1.02 report on original thought and memory. Github Repository: https://github.com/10111two/primite-1.02
You raise several good points, and I completely agree. Intelligence is an emergent property but there are some prerequisite to intelligence. Memory formations, original thought generation (imagination and creativity), anticipation, multifunctional trigger ability (think of how our one sense triggers responses of other senses). These emergent properties result in intelligence. We are exploring this principle first approach at JN Research. If you are interested, i would recommend this blog post: https://jn-research.com/blog/f/beyond-ai-%E2%80%93-introduci...
Yes. Current deep learning is powerful but fundamentally statistical. It classifies, predicts, and optimizes, but it doesn’t cognize. The assumption is that by scaling parameters and data, generalized intelligence will “emerge.” There is a big possibility that the fundamental framework of traditional AI is incomplete. It’s roots are inspired by Hebbian Learning, which itself is descriptive in nature. A good analogy is to think about Newton’s law of Gravity, works tremendously work, but it couldn’t explain the orbit of mercury. No matter how much we tweaked the maths, it didn’t fit. The framework was incomplete. It took Einstein’s general relativity – a new framework who eventually explained it. If the framework itself doesn’t allow for the truth we are after, we can spend trillions, it not gonna happen. It like having a design of motorcycle and somehow expecting it for fly. Currently, in AI landscape our efforts are being poured to build ramps. So that we can create an illusion that we have a flying machine. It can be argued that its also flying but claims are very misleading.
At JN Research, we took a principle-first route. We built Adaptrons - artificial neurons that actually exhibit graded potentials, subthreshold states, and action potentials. When networked, even small systems (hundreds to thousands of units) show memory formation, dreaming, anticipation, and original thought generation.
This is not ML, not symbolic AI. It’s a new substrate for cognition. If you’re interested, check us out here: https://jn-research.com
Few Anticipatory Questions
• “Isn’t this just ML/randomness?” - No training or gradient descent is used. Only neuron-like rules (graded, subthreshold, action potentials). Outputs are logged and timestamped; anyone can verify them.
• “How do you define ‘original thought’?” - An output is “original” if it was never presented as a stimulus during that system’s lifetime, yet emerges autonomously.
• “What about controls?” - We ran multiple experiments with different genetic parameters; each yielded different system behaviors. One run was deliberately configured as a pure input/output machine, confirming that adaptability is essential for higher functions.
• “Independent replication?” – We are open to live demos (reviewers choose inputs) and will provide full raw outputs. Under NDA, reviewers can also set genetic parameters and observe the system’s lifetime behavior.
• “Why 1,800 Adaptrons?” – Our approach is milestone-driven: we demonstrate emergence at small scales first (memory, dreams, anticipation), then scale gradually (20k, multimodal, 1M).
We know this is unconventional and expect skepticism. Our goal isn’t to make hype claims but to provide verifiable outputs, invite critique, and refine the framework. Happy to engage with specific test suggestions from the community.
At JN Research, we are exploring a third path between mainstream traditional AI and descriptive neuroscience. Instead of scaling or optimizing trained function approximators, we build Adaptrons; artificial neurons that behave like biological neurons (subthreshold + graded + Action Potential) and autonomously adapt internally and with other Adaptrons in a system. On this substrate, our small artificial brain Primite 1.03 (1,800 Adaptrons) now shows:
• Autonomous sleep states (no external input), with internal “dreams” and some shared as outputs.
• Original thoughts (novel images not seen as stimuli) arising during sleep and while awake.
• Memory formation and consolidation (short/intermediate/long-term), including memories of dreams later recalled while awake.
• Anticipation: outputs that appear before the corresponding stimulus is presented.
We ran 7 independent experiments with different genetic parameters and share detailed counts, timing, and example outputs. This is not ML training; it’s a principles-first cognitive substrate where higher functions emerge from the interaction rules. Furthermore, we also show that higher cognitive functions do not need bigger models or scale to emerge, we can see their early signs if the fundamental framework allows for it. If you are curious (or skeptical), we have included the full technical report and a data repo with outputs for verification, plus our prior 1.02 report on original thought and memory.
Github Repository: https://github.com/10111two/primite-1.03