hyzo

@hyzo

Study of crypto ecosystem + Ai + innovations + processes in science and our lives

181 following465 followers

Latest casts

hyzo@hyzo·
Reply

Investing in various other assets is an important component. Perhaps the digital-world equivalent of real estate is not as popular due to fears of hacking and loss even if it could be more profitable and beneficial in the long run or simply because saving for a physical home is considered more important than investing in digital assets. Perhaps you have thoughts on what might serve as a true equivalent?

hyzo@hyzo·

So, what does a well actually do? The depth and shape of the well determine the frequency of the particle's oscillations, which manifest in optical spectra. The very existence of such potential wells is crucial; they restrict diffusion between particles. *Diffusion is essentially a form of mixing it occurs when the molecules or atoms of one substance penetrate into the molecules or atoms of another substance.

hyzo@hyzo·

Potential well. And what is such a well actually for? A potential well is a finite region of space where the potential energy of a particle is lower than it is outside that region. *Potential energy is the energy a particle acquires due to its position relative to other particles, as well as due to internal stress within the particle or object, or due to electric charge, etc. A well typically has a width and a depth (or height depending on how you look at it, heh 🙃). ~ The point with the lowest potential energy value is called the bottom of the well. ~ Points where the total energy is less than the height of the potential well are called turning points. If a particle’s total energy is greater than the height (or depth) of the well, the particle will oscillate within it. To escape the well, the particle must gain energy. https://en.wikipedia.org/wiki/Potential_well

hyzo@hyzo·
Reply

thanks for the statistics and thoughts!, how do you like the latest updates in fileverse?, should be user-friendly

hyzo@hyzo·

In short, this is a highly valuable tool. It allows for the verification of security within the quantum realm; for instance, in quantum cryptography, it can determine whether a protocol inadvertently reveals messages through quantum data accessible to an adversary. https://arxiv.org/abs/2611.14147

hyzo@hyzo·

Consensus: Validators evaluate these "masks" and vote on the best options. Iterative Cycle: An intermediate layer of the neural network is used to randomly verify the work of validators (specifically, to check whether their updates are based on valid data). Important clarification: While ZK proofs in this context verify the correctness of a specific computation, they do not prevent the submission of falsified data; the fourth mechanism (the iterative cycle) is designed to address this. This is an interesting area of ​​research aimed at enhancing participant privacy during AI model training. Naturally, there are some drawbacks—such as the lack of complete privacy due to the use of intermediate neural network layers, or the issue of data reuse.

hyzo@hyzo·

Decentralization + ZK + blockchain + AI training. This approach combines cutting-edge fields to facilitate AI training. It proposes a method for collaborative AI training within a decentralized network, utilizing Zero-Knowledge (ZK) proofs to ensure privacy. The challenge with conventional methods is that participants must disclose their data; furthermore, they might possess low-quality data or act dishonestly. The proposed approach addresses these issues through several mechanisms: Verification: A participant runs a consensus-agreed AI model on their own data and generates a ZK proof confirming the correct execution of the computation, which the network then verifies. Filtering: Validators analyze the gradients used to update model parameters and create "masks." In this context, a mask acts as a quality control mechanism, specifying which parameters to retain fully and which to attenuate. arxiv.org/abs/2611.11477

hyzo@hyzo·

Lattices A paradise for mathematicians, and a nightmare for quantum computers. The topic of post-quantum cryptography is currently of critical importance. If we wish to protect assets and information, we must find ways to do so (we will focus specifically on modern cryptography). So, let’s break down what this is, why it is needed, and how it works. A lattice is an infinite set of points in a multidimensional space. This set of points is formed by linear combinations of basis vectors using integer coefficients (essentially, the sum of vectors lying along different lines, multiplied by integer coefficients). The security of the encryption relies on specific mathematical problems—specifically, problems that are computationally infeasible to solve quickly within a multidimensional space, such as the Shortest Vector Problem (SVP) and Learning With Errors (LWE).

hyzo@hyzo·

Logical verification of models involving mutually exclusive options. Sometimes, models may fail to ensure that the sum of probabilities equals 100%. Testing was conducted using simulations of social networks, real-world statements, and scenarios with two mutually exclusive outcomes. Research findings: Wording matters. Errors can propagate through a network of agents. Mathematical correction helps, though it does not always improve accuracy. Logic is crucial—especially in chains involving multiple agents; if each agent alters their assessment with every rephrasing, the final result will be entirely different. https://arxiv.org/html/2610.32874v1

hyzo@hyzo·

Communication between AI agents. Let’s examine the key mechanisms: A2A: An open standard that allows different AI agents to discover one another on a network, exchange information and results, and assign tasks. JSON/API: Messages are transmitted in a structured format via JSON requests or API calls with clearly defined parameters for both requests and responses. Model Context Protocol (MCP): An open standard that facilitates communication between agents within shared tools (without exposing the context or data from other systems). Multi-agent system architecture represents another approach to communication: Lead/Manager AI: A primary AI receives a task from a human, analyzes it, and distributes sub-tasks among other relevant agents. Specialized Agents: You are likely aware that certain agents have specific areas of expertise—meaning they excel at particular types of tasks (such as logical/mathematical calculations, graphics, or video processing). These can be classified as specialized agents. Shared Memory: A common database where agents can write and retrieve data; it also serves as a space to leave messages or intermediate results for other agents to access.