I did a lot of early neural net research (e.g. DNNs in 1986, Real Time Recurrent Learning in 1987) and then applied to speech recognition (1991). I’m now researching new architectures.