My thoughts on ChatGPT & Language Model evolution

 I have been anxiously following the ChatGPT hype cycle lately like many others. I love how Silicon Valley can hype  a technology and seemingly consumes the public discourse overnight. I've been working with language models since 2017 and I can't deny the power of them includung ChatGPT.  In many ways, a general purpose and large scale language model is intriguing and slap a conversartional chat interface on it and you lower the barrier of entry for millions of people. At the same time, you create powerful feedback loops.  I have my doubts that we are approaching a singualrity event where one general purpose model takes over the world. In contrast, I bet that domain specfic and more specialized language models will still dominate and outperform general purpose in their respective domains for quite some time.  In parallel, the cost of building a large scale language model like ChatGPT is  being commoditized and the price will collapse. Differentiation will likely shift to the proprietary data you have access to differentiate one model from another. 

I read the Google Bard blog announcement from the CEO. It seemed reactive and they are moving too conservatively than the Google of the past.