The gradient descent of knowledge
Deep Prasad (yug-cybera) 🏴☠️
Verified account This account is verified. Learn more@Deepneuron
Creating the stack for multi-species intelligence. Physics-native AGPI + artificial life + silicon life + whatever is listening.🛰️📡StarVasa CEO 🏴☠️
1.7K Following 34K Followers 69K Posts
Latest posts 5
A small animation on the type of neural networks we’re developing, they’re self-repairing, unlike the neural networks of today.
NNs are what power ChatGPT and modern AI today
NNs are what power ChatGPT and modern AI today
0:40
With the help of GPT-6 Astra and Claude 5.2, I think I might have discovered a new class of AI systems!
The idea is creating artificial astrocytes that mimic the behavior of glial cells (something Einstein had significantly more of in his brain than the average person).
The idea is creating artificial astrocytes that mimic the behavior of glial cells (something Einstein had significantly more of in his brain than the average person).
0:43
Building something past ASI: AGPI, Artificial General Physics Intelligence. A physics-native superintelligence that can rewrite itself from first principles, governed by rules it can't rewrite. In other words, it’s an Intelligence that doesn't just know more and knows the why.👽
Had a fascinating conversation today about spacecraft flight readiness left me thinking about a problem that will become increasingly important as AI enters mission control.
Who validates the decisions of autonomous systems?
At StarVasa, our patented mission-control architecture translates high-level objectives into spacecraft subsystem commands.
But generating commands is only part of the challenge.
The harder question is whether those commands respect physical constraints, whether their assumptions are supported by evidence, and whether they can be validated before execution.
Simulation, hardware-in-the-loop testing, and independent decision assurance all have roles to play.
We’re going to need referees for machine intelligence in space.
Because intelligence can propose.
Physics gets a veto. 🛸
Who validates the decisions of autonomous systems?
At StarVasa, our patented mission-control architecture translates high-level objectives into spacecraft subsystem commands.
But generating commands is only part of the challenge.
The harder question is whether those commands respect physical constraints, whether their assumptions are supported by evidence, and whether they can be validated before execution.
Simulation, hardware-in-the-loop testing, and independent decision assurance all have roles to play.
We’re going to need referees for machine intelligence in space.
Because intelligence can propose.
Physics gets a veto. 🛸