Federated learning (FL) enables collaborative model training across distributed clients while preserving data locality. Although FedAvg pioneered synchronous rounds for global model averaging, slower devices can delay collective progress. Asynchronous FL (e.g., FedAsync) addresses stragglers by continuously integrating client updates, yet naive implementations risk client drift due to non-IID data and stale contributions. Some Blockchain-based FL approaches (e.g., BRAIN) employ robust weighting or scoring of updates to resist malicious or misaligned proposals. However, performance drops can still persist under severe data heterogeneity or high staleness, and synchronization overhead has emerged as a new concern due to its aggregator-free architectures. We introduce Fast-and-Reliable AI Network, FRAIN, a new asynchronous FL method that mitigates these limitations by incorporating two key ideas. First, our FastSync strategy eliminates the need to replay past model versions, enabling newcomers and infrequent participants to efficiently approximate the global model. Second, we adopt spherical linear interpolation (SLERP) when merging parameters, preserving models' directions and alleviating destructive interference from divergent local training. Experiments with a CNN image-classification model and a Transformer-based language model demonstrate that FRAIN achieves more stable and robust convergence than FedAvg, FedAsync, and BRAIN, especially under harsh environments: non-IID data distributions, networks that experience delays and require frequent re-synchronization, and the presence of malicious nodes.

Multi-level optimization has gained increasing attention in recent years, as it provides a powerful framework for solving complex optimization problems that arise in many fields, such as meta-learning, multi-player games, reinforcement learning, and nested composition optimization. In this paper, we study the problem of distributed multi-level optimization over a network, where agents can only communicate with their immediate neighbors. This setting is motivated by the need for distributed optimization in large-scale systems, where centralized optimization may not be practical or feasible. To address this problem, we propose a novel gossip-based distributed multi-level optimization algorithm that enables networked agents to solve optimization problems at different levels in a single timescale and share information through network propagation. Our algorithm achieves optimal sample complexity, scaling linearly with the network size, and demonstrates state-of-the-art performance on various applications, including hyper-parameter tuning, decentralized reinforcement learning, and risk-averse optimization.

SingularityNET's founding mission of creating beneficial decentralized AGI is a large one and has a number of different aspects, all of which need to work together tightly. There's the decentralized-platform software layer, which is needed to make it feasible to run AI systems across large numbers of computers without any central owner or controller. The SingularityNET protocol and platform provide one portion of this layer, the NuNet decentralized compute resource network provides another, and the HyperCycle AI-customized ledgerless blockchain provides another. Decentralized-AI-platform tools are mainly agnostic in regards to how AGI is achieved -- be it neural nets, artificial life systems, brain simulations, logical reasoning engines, emergent combinations of different AI agents and paradigms, or something new and yet unforeseen. However, our team at SingularityNET has been putting significant effort into one particular approach to AGI, which is the OpenCog Hyperon framework, bringing together neural nets, logical reasoning, evolutionary learning, and other methods in the context of a distributed self-modifying knowledge metagraph.

Personalization in federated learning (FL) functions as a coordinator for clients with high variance in data or behavior. Ensuring the convergence of these clients' models relies on how closely users collaborate with those with similar patterns or preferences. However, it is generally challenging to quantify similarity under limited knowledge about other users' models given to users in a decentralized network. To cope with this issue, we propose a personalized and fully decentralized FL algorithm, leveraging knowledge distillation techniques to empower each device so as to discern statistical distances between local models. Each client device can enhance its performance without sharing local data by estimating the similarity between two intermediate outputs from feeding local samples as in knowledge distillation. Our empirical studies demonstrate that the proposed algorithm improves the test accuracy of clients in fewer iterations under highly non-independent and identically distributed (non-i.i.d.) data distributions and is beneficial to agents with small datasets, even without the need for a central server.

Join us on November 9 to learn how to successfully innovate and achieve efficiency by upskilling and scaling citizen developers at the Low-Code/No-Code Summit. All of these are films where machines become sentient and attempt to take over the world (or at least kill all humans). It's a popular plot line because it speaks to our deep-seated fears about technology. Will our devices and the data they collect be used against us as we move toward Web3? In recent years, we've seen increasing evidence that our data is being used in ways we never intended or anticipated.

I recently started an AI-focused educational newsletter, that already has over 70,000 subscribers. TheSequence is a no-BS (meaning no hype, no news etc) ML-oriented newsletter that takes 5 minutes to read. The goal is to keep you up to date with machine learning projects, research papers and concepts. One of the pivotal challenges of the next decade of artificial intelligence(AI) is to determine whether data and intelligence are democratized or remain in control of a few large organizations. A few months ago, I wrote a three-part series of the decentralization of artificial intelligence(AI).

With the increased pressures placed on the business over the past year or so, perhaps it should come as little surprise that many enterprises chose to delay or are lagging behind in their digital transformation pushes. But simultaneously, organizations are acknowledging that a digital overhaul is now a prerequisite for continued growth, with artificial intelligence (AI) and blockchain technologies establishing themselves as two powerful DX trends in recent years. Long the stuff of science fiction fodder, AI has often been depicted as running amok and attempting hostile takeovers of electronic systems in media. But recent Google research indicates AI can become overly aggressive in the right circumstances, or is that the wrong circumstances? Either way, the intrinsic structural elements of another major enterprise trend, blockchain tech, could be the solution to combating AI aggression.

One of the pivotal challenges of the next decade of artificial intelligence(AI) is to determine whether data and intelligence are democratized or remain in control of a few large organizations. Last year, I wrote a three-part series of the decentralization of artificial intelligence(AI). In that essay, I tried to cover the main elements that justify the movement of decentralized AI ranging from economic factors to technology enablers as well as the first generation of technologies that are developing decentralized AI platforms. The arguments made in those articles were fundamentally theoretical because, as we all know, the fact remains that AI today is completely centralized. However, as I work more in real world AI problems, I am starting to realize that centralization is an aspect that is constantly hindering the progress of AI solutions.

In 2014, the Society of Automotive Engineers (SAE) classified the future of automobile in 6 levels (from 0 to 5). Car makers have managed to develop self-driving cars stepping into level 3 (conditional automation: autonomous under certain conditions) and soon challenging level 4 (high automation: the driver can take control of the car if necessary) and 5 (full automation: the driver is not required at all) using a combination of cutting edge technologies such as captors, sensors, radars, and other visual systems used for mapping, localization, and obstacle avoidance. Today, sensors such as LIDAR, GPS, or IMU are used to gather information around the car that will be processed for the car to react accordingly. These technologies have allowed companies to push self-driving cars to the market but self-driving cars are yet still in the testing phase and there are major challenges to overcome before thinking of reaching full autonomy. As of today, self-driving cars still have challenges to tackle concerning communication and security.

"There is no spoon," says the little boy in The Matrix. "… you'll see that it is not the spoon that bends, but only yourself." Armed with that nugget of wisdom, Neo begins his journey to change the nature of reality, by changing himself. When people have tried to explain the concept of blockchain, they have often stated that it allows for the decentralization of organizations and that it has the technological power to remove the centralized party. In other words, in blockchain-enabled networks: there is no decision-maker, no one to lead, no one to request access from, no one to control and no one to steer the evolution of these networks.