The Yegnanarayana Problem

Bayya Yegnanarayana (“Yeggi”) was a professor of Computer Science at IIT Madras. Among other things, he taught a course on Speech Technology that I happened to take in 2003. He taught us well, and I learnt a lot in that course, but there is one thing that I remember from it.

We developed the technology to recognise speech (artificial neural networks and a precursor to deep learning, I realise in hindsight) all those years ago, said Yeggi, but people told us then that our computers are not fast enough to recognise speech. It has been so many years since then, and computers are an order of magnitude faster than they were when this technology was developed, but we are no closer to getting computers to recognise speech, he went on. He also added that leave alone recognising speech, we can’t even get computers to convert text to speech.

A year later, I graduated. A year after that, Yeggi retired from IITM and moved to IIIT Hyderabad. Computers continued to become faster. And then less than five years later, computers became fast enough that deep learning became a thing. Now deep learning is everywhere, and speech recognition is so commonplace that leave alone computers, most smartphones can support it.

I don’t know what Yeggi tells his students at IIIT nowadays.

Now, I don’t intend to pick on Yeggi. He was a wonderful teacher, and I learnt a lot in that speech technology course. However, I want to associate his name to the kind of lament that comes just before something hits a point of inflexion, where there has been slow and steady progress for a long time, but without bearing fruit (yet), and then suddenly progress happens.

This is not limited to technology alone. It happens everywhere. You want to get somewhere and you come up with a process to get there. You diligently follow the process but nothing happens, for a while. And then sometimes, rather than keeping at it, you give up, and then you see progress would have happened.

Inflexion points are everywhere. The problem is that you don’t know when they will occur. The best you can do is to get together as a group and follow the process.

Put Comment