A distributed artificial intelligence (DAI) system is a type of artificial intelligence (AI) solution that uses blockchain technology to distribute, process, and store data across a network of nodes.
Traditionally, the development of AI systems has been siled among a small number of technology vendors, such as Google and OpenAI, who have had the funding needed to develop the infrastructure and resources needed to build and process large datasets. I did.
However, the centralization of AI development in the industry requires significant funding to enable organizations to develop and process the data they need to compete in the marketplace.
For example, the main bottleneck we see in AI today is the lack of GPUs. Deep learning, the artificial intelligence (AI) technique behind large-scale language models (LLMs) like GPT, is a long-running, compute-intensive process at scale. The more parameters an LLM has, the more GPU memory it requires to operate.
Although it still has some drawbacks, Web3 infrastructure has the potential to address the challenges posed by AI integration and provides opportunities for innovative solutions, as discussed below.
Distributed AI computing network
Distributed computing networks connect individuals who need computing resources with systems with unused computing capacity. This model allows the network to offer more cost-effective pricing as opposed to centralized providers, as individuals and organizations can contribute idle resources to the network at no additional cost.
Distributed GPU rendering facilitated by blockchain-based peer-to-peer networks has the potential to scale AI-powered 3D content creation in Web3 games. However, a significant drawback of distributed computing networks is that communication overhead between different computing devices can slow down machine learning training.
Decentralized AI data
Training data serves as the initial dataset used to teach a machine learning application to recognize patterns or meet certain criteria. On the other hand, test or validation data is used to evaluate the accuracy of the model, but since the model is already familiar with the training data, another dataset is required for validation.
There are ongoing efforts to create a marketplace of AI data sources and AI data labeling, with blockchain serving as an incentive layer for large enterprises and institutions to improve efficiency. However, at their current early stages of development, these industries face hurdles such as the need for human review and concerns surrounding blockchain-enabled data.
For example, there are service provider (SP) computing networks designed specifically for training ML models. SP computing networks are tailored to specific use cases and typically employ an architecture that consolidates computing resources into a unified pool, similar to a supercomputer.
The SP computing network determines costs through community-controlled gas mechanisms or parameters.
distributed prompts
Full decentralization of the LLM presents challenges, but the project is exploring ways to decentralize prompts by encouraging the contribution of self-trained techniques. This approach incentivizes creators to create content and provides an economic incentive structure for more participants in the landscape.
Early examples include AI-powered chatbot platforms that tokenized incentives for content creators and AI modelers to train chatbots. Chatbots can then become tradable NFTs that grant access to user-granted data to train and fine-tune models. The decentralized PromptHis Marketplace, on the other hand, aims to incentivize prompt creators by providing ownership of their data and allowing their prompts to be traded on the marketplace.
Below we take a look at some applications of decentralized artificial intelligence in the market today.
- bitensor We aim to revolutionize the development of machine learning platforms and create a neural internet. The project is establishing a peer-to-peer marketplace of machine intelligence where AI models can combine their intelligence to essentially create a “digital hive mind.” This innovative decentralized approach is designed to enable rapid scaling and knowledge sharing between AI systems.
- Singularity NET is a blockchain platform that allows anyone to build, share, and monetize AI services. There is an internal marketplace where users can browse and pay for AI services in the platform's native cryptocurrency AGIX. Developers can monetize their AI solutions and models without having to fully build and develop apps for end users. Similarly, developers can purchase AI solutions and models to use in their applications.
See also
- ocean protocol is an Ethereum-based platform that allows businesses and individuals to exchange and monetize data and database services. This may include making data available to researchers and start-ups without the data owners relinquishing it.
Web3 offers a promising path towards a more inclusive future for artificial intelligence. Decentralized solutions enable individuals and organizations of all sizes to participate in and benefit from AI innovation by democratizing access to computing resources, data, and AI development tools.
Although challenges remain, continued advances and initiatives such as decentralized AI computing networks, AI data marketplaces, and collaborative AI models point to a transformative shift toward a more just and accessible AI environment. Masu.
______________________________________________________
related
