For the purpose of this blog the term AI refers to a traditional text book (Russell & Norvig) definition of it applying to the combination of Computer Science+ Statistics + (Language/Vision/Robotics). In IT —compared to most other industries— the time gap between research and practice has traditionally been very small.
For those working in the field of Data Science & AI the last few few years —especially the last 12 months —has seen incremental innovation like never before. (The seminal work in AI has arguably happened pre-2019: think of Convolutional Neural Networks, Word Embeddings, Self-Attention.) The length of some newsletters tracking these developments is substantially increasing (tripling/quadrupling) too. Think about what would be the shelf life of a learning-by-doing book on GenAI. Rarely, would it exceed an year. In the Enterprise world: in the implementation of the latest innovations for real world business problems, the risk of getting obsolete by the time the project completes a few months in production is very real! What this means is that the Innovation in AI could soon reach an escape velocity. Let me explain. This would mean a new State of the Art (SoTA) is available before the previous SoTA is used in practice. While this may be good for innovation for innovation’s sake it would mean that there will now be a a growing time gap between research and practice.
This would mean—that much like other industries—an understanding of the growing time gap between research and practice is required. Moreover, as IT flows through all industries it means this needs a wider appreciation. Also, continuous incremental innovations mean that the gap may not get narrower anytime soon. Unreasonable expectations could lead to disappointment! What do you think?
Aniruddha M Godbole is an inter-disciplinary expert. He is a continuous learner. These are his personal views.