My name is Malcolm, co-founder of ITOKA. We were one of the pioneers in building AI projects on the ICP as early as 2022.(I believe we are older than ICircle in proposing the AIGC concept for NFT). Having been deeply involved in the IC community for years and engaged with the community, I would like to share my perspective as a machine learning algorithm developer who is a true user of EMC
I first learned about EMC’s integration with ICP in March-April through Herbert’s Twitter post. Initially, the concept of an L2 solution on ICP didn’t immediately catch my attention. Like many others, I believed IC was primarily designed for computation, and I couldn’t understand what a dummy is building a layer-2 to slightly boost a 200mph Lambo to 201 
. However, when I discovered the purpose of EMC, it immediately piqued my interest, and I couldn’t wait to learn more about the team behind it. What they were doing aligned perfectly with my own aspirations. Having previously mined Ether using GPUs, which are also used for machine learning training, I wondered if we could leverage Ether miners to train AI models and reward them with tokens. Such an approach would benefit both the crypto and AI communities. Unfortunately, due to my limited knowledge of hardware, I wasn’t sure where to start, so I focused my efforts on algorithm development.
In late 2022, when Ethereum 2.0 transitioned to PoS, many miners faced difficulties as the network no longer required extensive GPU resources. Around the same time, ChatGPT emerged and showcased the power of using hundreds of Nvidia A100 GPUs to fuel smart AI applications. The discussion within the crypto space grew increasingly intense, with some individuals abandoning their careers in favor of AI to chase the GPT hype. However, those of us who were already involved in both AI and blockchain quickly realized that merging the two technologies could create a trillion-dollar business opportunity. What was once an aspiration had now become an urgent necessity.
During ETHDenver, I had the opportunity to meet Jan, the CTO of Dfinity, and share my idea about the potential of on-chain AI. I proposed the idea of leveraging GPUs to support parallel computation for AI on ICP and post a new consensus mechanism to replace number guessing in forum. Jan encouraged me to give it a try, and I eagerly embraced the challenge. However, as a machine learning engineer, designing a distributed computation network was daunting due to my limited hardware expertise (again).
In summary, building an on-chain AI infrastructure is tremendously challenging and requires individuals with a deep understanding of machine learning systems, distributed computation, blockchain consensus, and a strong drive to achieve this objective. After speaking with Zed, I am confident that his team possesses these qualities and that EMC will effectively address the challenges I have raised.
Zed is an experienced engineer who has developed large-scale applications for enterprises for over a decade. When I shared my concerns about niche aspects of AI, such as how EMC covers inference versus training, he quickly grasped the essence of my concerns and explained them to me clearly. This solidified my belief that EMC, under his leadership, would be a game-changer in addressing these challenges.
Our team is overwhelmed to develop our own music AI algorithm, but I am eagerly awaiting the opportunity to test EMC’s testnet and run ITOKA’s AI on their network. The ability to make our AI fully on-chain while reducing costs compared to services like AWS is incredibly exciting. Although I am not an investor and cannot assess the potential for a project’s financial success, I will be deeply disappointed if EMC fails. I look forward to witnessing the success of this project.