How NEAR is formed
A little over a year ago, NEAR.ai was founded by Alexander Skidanov and a colleague, with a strong conviction that program synthesis, a field delving into automated programming from human specifications, held immense potential to revolutionize software development and beyond.
Their belief stemmed from witnessing various research papers at ICLR 2017, which ignited their excitement about bringing program synthesis technology to the forefront. Having dabbled in program synthesis during their university days, they recognized the challenges posed by the nascent stage of machine learning at that time.
Post-conference, they envisioned NEAR.ai and identified the urgent need to establish a benchmark for the community. They drew parallels with the transformative impact of the ImageNet benchmark on Computer Vision and aimed to create a complex dataset that would both challenge the community and guide research.
Their initial focus was on programming competitions, as these provided a rich source of data for machine learning, comprising problem descriptions and corresponding solutions. However, the complexity of task descriptions posed a significant challenge, necessitating crowd-sourced efforts to simplify them.
Despite their efforts, the results obtained from their dataset, as presented in their ICML 2018 paper, fell short of practical usability. Concurrently, they engaged with potential users to gauge market demand, developing prototypes and testing user experience models.
Realizing the limitations of machine learning, particularly in natural language understanding, they explored alternative avenues, including the field of generating smart contracts for Ethereum. This led to a deeper understanding of blockchain technology and its potential to address existing challenges in software development and data privacy.
