Oops! We couldn’t find any results...
Republic Runtime is thrilled to announce our position as an institutional staker for Masa protocol alongside their new Bittensor subnet.
About Republic Runtime:
Republic Runtime operates an institutional staking platform that caters towards base-layer protocols, treasuries, and asset managers alike. Republic Runtime is proud to announce their collaboration with Masa as an initial institutional staking collaborator.
Financial Opportunities in AI:
The AI boom has brought out a tremendous amount of capital and funding, with early projects raising capital around ChatGPT wrappers, a worthy but ultimately misplaced allocation.
As the froth has settled the industry has come to terms with the fact that the value lies in the primitives, that is, the ‘large language models’ (LLMs) and proprietary modeling systems that convert massive data troves into intelligent output. These systems operate across a variety of ‘modalities’ or differentiating inputs like text, imagery, sound, etc… all capable of outputting like-content.
We are now seeing large technology corporations rallying around these primitives operators, like OpenAI, in order to integrate these new functionalities into the product suite.
According to AI Index, the global AI market revenue for 2023 was approximately $433 billion. This growth was driven by increased adoption of AI technologies across various sectors, including generative AI, which saw significant investment and integration into business operations
Looking ahead to 2024, the AI market is projected to continue its rapid expansion. Estimates suggest that the market size will reach $184 billion, with AI chips alone expected to generate $71 billion in revenue (AI Index).
As for big tech, a quick search reveals substantial profit:
Microsoft:
AI-driven cloud services, including Azure AI and Copilot features in Office products, significantly contributed to revenue.
Total revenue for fiscal year 2023: $211.9 billion (McKinsey & Company) (Investopedia).
Google (Alphabet):
AI-driven advertising tools, Google Cloud AI, and AI research projects like DeepMind were major revenue drivers.
Total revenue for 2023: $283 billion (McKinsey & Company) (Statista).
Amazon:
AI applications such as Alexa, recommendation algorithms, and AWS AI services contributed significantly to revenue.
Total revenue for 2023: $469.8 billion (Statista).
NVIDIA:
AI hardware, particularly GPUs used for AI and machine learning, saw massive revenue growth.
Total revenue for 2023: $26.97 billion (Investopedia).
Masa: The decentralized AI and data LLM network
“Masa is a peer-to-peer protocol designed to create a global, decentralizeddecentalized, and incentivized network for Fair AI”. -Masa Whitepaper
As the AI wave takes hold we are presented with the opportunity for new and amazing applications and innovations, but they require an abundance of specialized data to manufacture. While data is abundant, it’s not always accessible, readily-usable, or cost-effective to power applications that can transform our daily lives or business at a deeper level.
While LLM models present mind-blowing outputs they will need to insert themselves into every facet of our lives and perform specialized functions in our businesses in order to provide the best-in-class service that their developers stake claims around.
The blockchain industry has long raised flags around private data capture and ownership rights in the digital realm. If we are indeed going to let these machines into our lives as a means of betterment, we would be remiss to sacrifice data sovereignty and our own economic benefit for ease-of-use.
This is where Masa comes into play. Masa democratizes access to AI technology, ensuring fair compensation for data and compute contributors. Masa creates a collaborative ecosystem where individuals worldwide can contribute to and benefit from the rapid growth of AI.
It utilizes blockchain technology and rewards incentives where people can earn by contributing data and compute resources by running worker nodes that scrape, structure, and transform a wide-range of data sources. Their permissionless system allows AI developers to access and utilize diverse data sets anywhere in the world. Masa also provides access to fine tuned, open source Large Language models (LLLM), as well as providing data for anyone to train their own specialized LLM. Masa’s data scrapers range from social to public web data, while ensuring personal data remains private. Its decentralized and self-improving network automates data collection, validation, and processing, ensuring efficiency and transparency in the data pipeline.
Masa Network actors:
The Masa protocol involves several key network participants, each playing a crucial role in the ecosystem's functionality and sustainability.
1) Worker Nodes (Users): Users run worker nodes and stake MASA to earn passive rewards. They fulfill data requests by either fetching data in real time or scraping static data sets that are permanently stored on the network.
2) Validators: Validators secure the Masa Network. They run a Masa validator node by staking MASA tokens. They validate transactions, data contributions, AI outputs, and participate in consensus to maintain network integrity, and evaluate performance of worker nodes for distributing em. Validators compete for limited slots based on performance metrics (utility) and earn rewards for maintaining network security and data quality.
3) Developers (Oracle Nodes): Developers access diverse data and LLM services to build AI applications. Developers stake MASA tokens and pay dynamic fees (gas) in MASA to submit requests to the network. Available real-time and static data come from sources like Twitter, Discord, podcasts, and web scraping, including raw, structured, annotated, and vectorized data sets. Access is available through running a node or using Masa’s API.
4) Governance Participants: These are MASA token holders who participate in the governance of the protocol. They have voting rights on key decisions such as protocol upgrades, changes in economic parameters, and other significant actions that impact the ecosystem. Governance participants are incentivized through governance rewards, which promote active and engaged participation in the protocol's decision-making processes.
Each participant in the Masa protocol ecosystem is integral to its operation, contributing unique resources and skills. They are rewarded through a combination of MASA tokens, transaction fees, and other incentives, creating a balanced and sustainable network where all actors benefit from their contributions.
Bittensor: A Network of LLMs
Bittsensor seeks to attack the existing multi-modal development of standard, centralized AIs. Modality, in the world of AI, represents specific capabilities depending on what kind of data it has been trained on. In this way, an AI trained on text and chats can produce a human-like conversation, think customer service bots. Whereas an AI trained on images can be used to produce imagery.
The problem area here is that each modality consumes a massive amount of training data, and even more relevant, requires a basket of mega-computing that is intensive in both energy and time.
Bittensor seeks to establish itself as an alternative by using token incentives to scale a group of networks for each model. In this way you can think of it as a ‘network of networks’ or ‘subnets’ as they are known within the Bittensor ecosystem. The idea being that an image-oriented subnet can exist inside of Bittensor’s distributed compute machine.
This permits anyone building image-oriented LLMs to reap the rewards of Bittensor’s token incentive design. Subsequently a text-oriented LLM can also populate the network, taking advantage of Bittensor’s economic design whenever text-oriented requests come in.
At scale, the Bittensor network pushes an image-oriented request to such image-oriented subnets, with a similar occurrence happening to LLMs that are text-oriented, voice-oriented, etc… In this manner the Bittensor network can push requests directly to the specific subnet that is best suited to facilitate the demand, leading to faster performance across a large set of decentralized hardware.
Bittensor and its $TAO token sit a layer above, facilitating the coordination of information between its subnets and the distribution/utilization of the compute that powers the network. With all of its elements are held in place by $TAO rewards, the network's way of incentivizing subnets to grow, update, and maintain appropriate uptime.
Bittensor's network operators are integral to the decentralized AI network, performing roles as miners and validators.
Miners:
- Function: Provide computational power and machine intelligence to the network.
- Contribution: Train AI models and generate digital commodities through competitive processes within subnets.
- Reward: Earn TAO tokens based on the quality and performance of their computational contributions, as evaluated by the network's consensus mechanism.
Validators:
Function: Verify the authenticity and quality of contributions made by miners.
Process: Utilize advanced cryptographic methods and consensus algorithms to ensure the integrity of the data and models produced.
Reward: Receive TAO tokens for their critical role in maintaining the network’s reliability and security.
Consensus Mechanism:
Bittensor employs an on-chain consensus protocol to transparently and fairly distribute rewards, ensuring high standards and incentivizing continuous improvement in AI outputs.
This technical framework ensures a robust, scalable, and secure ecosystem for AI development. *Bittensor docs
Masa and Bittensor:
Masa just announced the release of their own Bittensor Subnet, which facilitates the decentralized gathering and utilization of data and compute to power LLMs and a variety of AI use cases. In this way it is leveraging the Bittensor network as a mechanism to incentive and reward Masa Network contributors. Together they represent a large part of the AI stack, decentralizing the data harvested for LLMs while also incorporating into a broader AI network for greater utility.
“As an early crypto builder and adopter who participated in Ethereum’s ICO in 2014, Bittensor reminds me of Ethereum’s ecosystem circa 2017,” said Brendan Playford, Co-founder of Masa.
"While Bittensor is still in its early days, it has the potential to surpass Ethereum’s growth, fueled by the rapid expansion of Decentralized AI. DeAI has the potential to become even bigger than Bitcoin. At Masa, we are integrating into the Bittensor ecosystem to exponentially accelerate the development of Decentralized AI, with data serving as the new currency of Fair AI.”
Conclusion:
There is no sign of slowing in either the capabilities of AI systems themselves or the revenue that they are able to capture. It is industry consensus that AI capabilities will not only be the foundation of society but also the leading revenue stream as humanity moves towards full-automation.
As such, these systems will require ever more input, consuming increasingly large data-sets and even building out new mechanisms for compressing and digesting large data. This will produce vast new products that facilitate an ease of life, remove frictions, and abstract the user interfaces into what are called ‘natural user interfaces’ (NUIs). But it also asks for something in return. The fundamental components of AI - compute, models, and data - should not be monopolized by centralized entities, generating trillions of dollars in value at the sacrifice of our data and privacy, with no economic benefit to the public whose data it's being trained on.
While “Bitcoin fixes this” may not be applicable here, the solution does indeed lie in blockchain and decentralized systems. A new AI paradigm is needed, where there is fair compensation for data contributors, democratized access to AI technology, and innovation unshackled by the constraints of centralized control.
Blockchains and distributed systems are the clearest answer to this issue and we are proud to support Masa in their mission to facilitate the building of a decentralized global data and compute network that rewards and protects users and entities alike.
*To learn more about Masa’s website and/or X account
