diff --git a/.github/spelling/known_words_corpus.txt b/.github/spelling/known_words_corpus.txt index a5fb0465a..988f2f977 100644 --- a/.github/spelling/known_words_corpus.txt +++ b/.github/spelling/known_words_corpus.txt @@ -767,3 +767,5 @@ responderagent geolocation chatbot vectara +tokenomics +incentivized \ No newline at end of file diff --git a/pages/concepts/_meta.json b/pages/concepts/_meta.json index 65b7dfe58..42ff2173b 100644 --- a/pages/concepts/_meta.json +++ b/pages/concepts/_meta.json @@ -5,5 +5,6 @@ }, "agent-services": { "title": "Agentverse", "timestamp": true }, "ai-engine": { "title": "AI Engine", "timestamp": true }, + "asi-train": { "title": "ASI", "timestamp": true }, "fetch-network": { "title": "Fetch Network", "timestamp": true } } diff --git a/pages/concepts/asi-train/_meta.json b/pages/concepts/asi-train/_meta.json new file mode 100644 index 000000000..9cdd1c68f --- /dev/null +++ b/pages/concepts/asi-train/_meta.json @@ -0,0 +1,6 @@ +{ + "asi-train-intro": { + "title": "ASI: introduction", + "timestamp": true + } +} diff --git a/pages/concepts/asi-train/asi-train-intro.mdx b/pages/concepts/asi-train/asi-train-intro.mdx new file mode 100644 index 000000000..5787f9bcf --- /dev/null +++ b/pages/concepts/asi-train/asi-train-intro.mdx @@ -0,0 +1,31 @@ +# Introducing ASI\ platform + +## What is ASI\? + +ASI\ is a decentralized platform designed to empower users to fund, train, and validate domain-specific AI models tailored for specialized industrial applications. Unlike general-purpose Large Language Models (LLMs) such as OpenAI's GPT or Meta's Llama—which are built for broad, general-purpose tasks — ASI\ focuses on enabling the development of precise and highly specialized AI solutions for well-defined industrial sectors such as robotics, biotechnology, molecular design, and drug discovery. + +The platform facilitates community-driven AI models development by allowing members to pledge resources or funding to specific projects. These contributions enable the training and deployment of models optimized for addressing the unique challenges faced in specialized fields. By leveraging Fetch.ai's blockchain technology, ASI\ ensures transparency, security, and decentralization, creating an ecosystem where individuals and organizations can actively contribute to and benefit from the advancement of industry-specific AI solutions. + +## Why ASI\? + +While LLMs have achieved remarkable advancements, their general-purpose nature limits their effectiveness in specialized fields. These models often lack the depth and precision required to address complex problems in domains such as healthcare, molecular design, and biotechnology. Their responses are too general or verbose, and they can fail to deliver the targeted insights that highly specialized fields demand. + +Additionally, because LLMs are trained on diverse datasets, they may inadvertently introduce biases that compromise outcomes in critical sectors where precision is paramount, such as medicine. Furthermore, their inability to handle multimodal data (e.g., integrating text, images, and other data formats) limits their applicability in industries that require a combination of inputs. +Scaling general-purpose LLMs to meet the unique demands of specialized industries is both impractical and unsustainable due to the immense computational resources required. The environmental and operational costs of training and deploying these models also make them unsuitable for many industrial settings. + +ASI\ overcomes these limitations by providing a decentralized platform that facilitates the development of domain-specific AI models. It ensures these models are optimized for accuracy, scalability, and energy efficiency, directly addressing the unique requirements of well-defined industries. + +## How does ASI\ work? + +ASI\ operates through a robust, multi-layered framework designed to ensure functionality, scalability, and efficiency. At its core are **Autonomous Inference Model (AIM) Agents**, which manage trained models and facilitate inferences for tasks in specialized domains, such as robotics, molecular design, or drug discovery. These agents act as intermediaries, wrapping the inference API to streamline interactions and deliver results efficiently. AIM Agents run on a distributed network of nodes, enabling fast, scalable, and decentralized AI inference. + +A network of validators verifies the outputs generated by AIM Agents and guarantee accuracy and reliability. Validators play a crucial role in maintaining the system's integrity and are rewarded with FET tokens for successful validations. By fostering trust and accountability, this decentralized validation mechanism ensures that the inferences delivered are of high quality and dependable. +The platform supports participation through a structured tokenomics model; AIM Agents earn FET tokens for processing inference requests, with additional bonuses for delivering accurate results quickly. Validators stake FET tokens as a commitment to unbiased assessments, earning rewards for correct validations while facing penalties for errors. Users pay for inference services based on task complexity, while network operators hosting the infrastructure are compensated in FET tokens to maintain system stability and scalability. + +The ASI\ framework is built on several interconnected layers. The **application layer** manages user interactions and payment processing, enabling easy access to the platform's services using FET tokens. The **resource management layer** dynamically allocates computational resources to optimize the use of network nodes for running AI models. The **validation layer** ensures the quality and accuracy of outputs through rigorous assessments performed by validators. The **blockchain layer** secures transactions, governs model updates, and maintains transparency and immutability across the platform. Together, these layers create a decentralized and efficient ecosystem for deploying and maintaining domain-specific AI models. + +FET tokens stand at the foundation of ASI\, driving transactions, rewards, and governance. Community members can pledge FET tokens to support the development of specific AI models, earning a proportional share of future revenues generated by these models. Longer pledges are incentivized with higher returns to encourage sustained engagement. Validators stake tokens to validate model outputs, earning rewards for accurate performance while being penalized for errors. Staking tokens also grants governance rights, allowing participants to influence platform decisions such as prioritizing models or determining revenue distribution. + +Revenue for the platform comes from multiple sources. Licensing agreements and intellectual property rights ensure fair compensation while fostering innovation. Direct inference APIs provide consumers with immediate access to model outputs, generating revenue through usage fees. Customized consultancy services enhance the platform's value by offering tailored solutions for specific industrial needs. Early participants are attracted with higher ownership stakes, reflecting the risks of supporting models during their development phase. As models mature and risks diminish, these incentives taper off. A dynamic pricing model ensures equitable resource allocation, while reinvested fees contribute to ongoing research, system sustainability, and environmentally friendly computing practices. + +By integrating AIM Agents that wrap inference APIs, ASI\ delivers a straightforward, scalable, and efficient platform for specialized AI-driven solutions. This comprehensive framework fosters collaboration, increases contributions, and creates a balanced ecosystem where stakeholders actively shape the development and success of domain-specific AI models.