Decide AI: The Internet Computer Future of Intelligence

The demand for Large Language Models (LLMs) is soaring, and the infrastructure to support this demand is crucial. Reliable models and high-quality data are essential to meet this need. Currently, the power and progress in LLMs are concentrated in centralized, closed-source, and general-purpose models. However, the long-term and most profitable opportunities lie in industries like health and finance that require tailored, high-performance models. There is a significant, largely untapped opportunity to create an environment for developing, refining, and collaborating on future-ready LLMs and supporting datasets.

The DecideAI Ecosystem

DecideAI is an ecosystem designed to meet the needs of the high-end, specialized LLM market. It consists of three products:

Decide Protocol

A platform that coordinates artificial and human intelligence to annotate, train, and continuously improve specialized LLMs and targeted datasets using Reinforced Learning with Human Feedback (RLHF).

Decide ID

A unique Proof of Personhood (PoP) methodology to verify and trace contributors and their credentials, ensuring high-quality data.

Decide Cortex

A platform offering access to pre-trained LLMs and pre-vetted datasets via direct purchase or API endpoints for clients and developers who do not want to start from scratch.

The DecideAI ecosystem is supported by a robust reward architecture that incentivizes long-term engagement and discourages bad actors.

The LLM Market

The LLM market has exploded in importance since 2022, and by 2030, it is expected to reach $260 billion. The main drivers of growth are practical applications in corporate or industrial settings, including healthcare, finance, e-commerce, and media. The demand for LLMs necessitates a corresponding supply chain. Raw data must be refined through labeling, annotating, and fine-tuning to be usable. Both artificial and human intelligence are valuable in this refining process.

General-Purpose vs. Targeted Models

LLMs can be divided into:

• General-purpose models (e.g., ChatGPT, Claude): Developed by large firms for mass use.
• Targeted models: Often developed by the companies that use them, focusing on quality rather than quantity.

General-purpose models require large datasets and workforces, while targeted models emphasize quality and domain-specific expertise. The latter represents the most attractive area for long-term growth in the LLM training market.

Challenges in the AI Industry

The current dominance of general-purpose LLMs by a few large, closed-source, centralized corporations creates several problems:

• Ethical concerns: Potential biases and inequitable outcomes.
• Privacy risks: User privacy can be compromised in pursuit of greater scale.
• Unreliable results: Large, generic datasets tend towards mediocrity and errors.
• Uncustomizable: Closed-source models cannot be easily adapted or fine-tuned.
• Slow to adapt: Dynamic fields require domain-specific expertise and context-dependent judgment.
• Scaling problems: High usage costs and risk of outages for enterprises with substantial demand.

Seizing the Opportunity: DecideAI

As AI's capabilities grow, so does the potential for misuse. DecideAI aims to counterbalance this by creating open-source, transparent, and secure AI infrastructure that protects user privacy while rewarding data contributions.

The DecideAI Ecosystem

The initial iteration of the DecideAI Ecosystem consists of three components, each addressing a key need in the AI market:

Decide Protocol: Transparent Training

A platform for sourcing and refining LLM training data, building new models, and compensating contributors.

Decide ID: Data Quality

A system for verifying the contributor workforce to ensure high-quality training data.

Decide Cortex: Seamless Collaboration

An open-source community for hosting and accessing models and high-quality datasets.

Contributors (data labelers, developers, etc.) are rewarded via the native token, creating a sustainable system where everyone benefits.

Tokenomics

Purpose of the DCD Utility Token

The DecideAI token (DCD) is the native utility token of the DecideAI platform, intended to:

• Reward participation
• Fund innovation
• Fuel collaboration

DCD tokens streamline activities across the platform, reward data creators and labelers, and incentivize developers and market creation.

Total Supply Levers

The total supply of DCD tokens at genesis is 1 billion. The supply will increase if more tokens are minted and decrease if tokens are burned. The DecideAI DAO uses a proposal system to manage token supply and rewards.

Incoming and Outgoings

At genesis, the SNS (Service Neuron System) will have a reserve of ICP (Internet Computer Protocol) and 426.5 million MOD tokens. The SNS will receive DCD tokens from partner applications and use them to reward users and for community bounties. Over time, the expectation is that incoming tokens will balance outgoings, allowing for higher user rewards and bounties.

Initial Token Allocation

The SNS was initialized with 1 billion tokens distributed across various categories, with different lock-up and release schedules to manage liquidity and incentivize long-term engagement.

Liquidity Management

Various tokenomics parameters can influence the total and liquid supply of DCD tokens. Initial values provide a balance of incentives, but these parameters can be adjusted by the DAO to manage supply and price.

Roadmap

• Q1 2024: Finalize product concept, initiate development, and begin blockchain integration.
• Q2 2024: Rebrand to DecideAI, launch Decide ID alpha, build backend infrastructure, and start partnership conversations.
• Q3 2024: Closed beta launch, token listing on CEXs, integrate Decide Protocol and Decide ID, and finalize computation partnerships.
• Q4 2024: Train and test flagship AI model, prepare for go-to-market.
• Q1 2025 and Beyond: Public launch of Decide Cortex, initiate post-launch support, and build the DecideAI ecosystem.

Team

• Raheel, CEO and Founder: Over a decade of experience in tech, University of Waterloo Software Engineering graduate.
• Pema, COO: Experienced public affairs and government relations advisor, master's degree from the University of Toronto.
• Jeshli, Lead AI Engineer: PhD in AI and Machine Learning, expertise in machine-learning algorithms and distributed systems.
• Vitaliy, Lead Software Architect: Over 10 years of experience, technical lead at many companies.
• Justin, Senior Full Stack Engineer: Experience in merchant onboarding and KYC.
• Alex, Front End Engineer: Passionate about design and creating functional products.

Conclusion

In conclusion, the future of specialized LLMs holds immense promise, particularly in sectors like healthcare and finance where tailored, high-performance models are increasingly indispensable. The DecideAI ecosystem emerges as a pioneering initiative poised to address the challenges of centralized, closed-source LLM dominance by fostering open-source collaboration, transparency, and data security. By leveraging the Decide Protocol for transparent training, ensuring data quality with Decide ID, and facilitating seamless collaboration through Decide Cortex, DecideAI not only aims to meet the growing demand for specialized LLMs but also to establish a sustainable and equitable AI infrastructure. With its innovative tokenomics and committed team, DecideAI is set to play a pivotal role in shaping the next frontier of artificial intelligence.