Artificial intelligence has made significant strides in recent years, particularly with the rise of generative AI and large language models (LLMs). These innovations have sparked our collective imagination, turning once fictional concepts into tangible products used by companies at scale. However, the process of training these models is both resource-intensive and expensive. Traditional cloud service providers can charge thousands of dollars per day for a single server instance, and demand for machine learning is rapidly increasing. In fact, the Machine Learning as a Service (MLaaS) market was valued at $5.7 billion in 2022 and is projected to grow to $31 billion in the next five years.
Interestingly, the computing power used for mining cryptocurrencies is more than 2.5 times larger than that of AWS, Google Cloud, and Azure combined. This begs the question: what if cryptocurrency mining could be used to train machine learning models?
Tromero has developed an innovative Proof of Useful Work consensus mechanism that allows distributed, decentralized computers to earn block rewards and fees by successfully training machine learning models. Unlike traditional proof of work systems such as Bitcoin, which rely on computationally expensive math problems for miners to solve in order to secure the network, Tromero’s approach replaces this process with useful machine learning model training.
Overall, Tromero’s technology offers a promising solution to the resource-intensive and expensive process of training machine learning models. By using their Proof of Useful Work consensus mechanism, companies can reduce computing costs and even generate revenue by contributing their computing power to train models.
To anyone watching technology news, it’s clear that we are currently in a (well deserved) bubble of hype and excitement around artificial intelligence. But, regardless of how the immediate crop of projects turns out, it is certain that the demand for training and fine-tuning ML models will continue to rise, all while compute costs remain high.
Tromero provides a solution to these cost and supply pressures by supplying compute providers and miners a stream of revenue to optimize their return on capital. Given the industry’s inability to keep up with demand for machine learning hardware, a compute network using crypto-economic incentives to drive utilization should result in a more balanced, democratized, and efficient network of machine learning resources.