More recently, NLP has been dominated by companies like Google, Facebook, Baidu, and Amazon that successfully create an intellectual property commercialization pipeline leading from academic research to production implementation at scale. Developing proprietary IP in NLP as a startup requires a programmatic approach to commercializing academic research.
Soffos is invested in building out its own commercialization pipeline for academic research in the following ways:
Our programs are global in scope, whereas there are some who might say that Silicon Valley focuses primarily on hiring alumni of a few select programs (Stanford, MIT, Carnegie Mellon).
As such, we are creating a sizable research community conversant with the best research happening in Big Tech and leading academic NLP departments, but who can see beyond the ‘groupthink’ that is beginning to hinder the “small world” network of Silicon Valley NLP.
China pioneered this approach to NLP commercialization. iFlyTek was an obscure, under-funded startup two decades ago that now runs ML systems on Intel clusters that rival the sophistication of incumbent heavyweights in machine translation. A similar approach will give us a chance to bootstrap likewise, and build narrowly-focused NLP applications that the major players overlook in their multi-decade quest to achieve general AI.
Long-term general AI research is a marathon that only enterprises and academic institutions can run, but enterprises and academic institutions will continue to rely on the agile approach of sprinters, the startups, to prove concepts and test markets for immediate business use cases along the way.