Difference between revisions of "Probabilistic Computing"

Latest revision as of 14:15, 21 May 2025

Meme

Start with noise, end up with information.

Context

Given some configuration, is it possible to just let the computer run and see where it gets?

Consider Cellular Automata, there the beginning slate is completely blank until some pattern is injected at the automata just runs it course.

Solutions

• 2025-05-19 Prototype Computer Uses Noise to Its Advantage Thermodynamic computing isn't just a rebrand of probabilistic computing

This approach is a natural fit for certain scientific computing applications that already include randomness, such as Monte-Carlo simulations. It is also well suited for AI image generation algorithm stable diffusion, and a type of AI known as probabilistic AI. Surprisingly, it also appears to be well-suited for some linear algebra computations that are not inherently probabilistic. This makes the approach more broadly applicable to AI training.

“Now we see with AI that paradigm of CPUs and GPUs is being used, but it’s being used because it was there. There was nothing else. Say I found a gold mine. I want to basically dig it. Do I have a shovel? Or do I have a bulldozer? I have a shovel, just dig,” says Mohammad C. Bozchalui, the CEO and co-founder of Ludwig Computing. “We are saying this is a different world which requires a different tool.”

Probabilistic Computing

• Uses probability-based logic rather than deterministic binary operations.
• Employs P-bits, which fluctuate due to thermal energy, making them useful for AI and machine learning applications.
• Can be 100 million times more energy-efficient than traditional GPUs by utilizing noise for complex calculations.
• Ideal for Monte Carlo simulations, probabilistic AI, and neural networks.

Thermodynamic Computing

• Uses physical principles of thermodynamics to process information.
• Relies on natural energy minimization to solve problems, rather than traditional logic gates.
• Can be applied to linear algebra, AI training, and Gaussian sampling.
• Works by allowing components to reach equilibrium, where the final state represents the solution.

Programming a Thermodynamic Computer

1. Define the Problem – Express computations in terms of energy-based models.
2. Configure Circuit Parameters – Adjust inductance, capacitance, and control voltages to shape probability distributions.
3. While theoretically different initial conditions would not impact results, that's not the way nature works. Check out a few to be safe.
4. Run Simulations – Observe how the system naturally evolves toward solutions.
5. Optimize for Efficiency – Fine-tune parameters to improve accuracy and stability.

Analog Computing

Programming an analog computer is quite different from programming a digital one. Instead of writing code, you configure physical circuits to model mathematical equations. Here’s how it works:

1. Define the Problem – Identify the mathematical equations you need to solve.
2. Set Up the Circuit – Use operational amplifiers, resistors, capacitors, and integrators to represent the equations.
3. Configure Inputs & Outputs – Adjust voltages and currents to match initial conditions.
4. Connect Feedback Loops – Ensure proper signal flow for integration and differentiation.
5. Run the Simulation – Observe results on an oscilloscope or other measurement tools.
6. Adjust Parameters – Fine-tune components to optimize accuracy.

References