Many people who know me would classify me as a skeptic (some might say cynic ).
I have stayed away from the current AI debate and will continue to do so. I worked on early AI solutions (expert systems in medicine) in the 1970s and concluded that the knowledge domain (human health) is so vast and importantly hugely complex that it was way beyond the technology at the time and certainly the big problems would not be cracked within my lifetime. I was right and continue to believe that and plan to stay well away from AI hype!
But Professor Brian Lucey, whom I follow, pointed out how easy it is (with Google NotebookLM) to create a fairly decent podcast from even an academic document.
Ever the doubter, I decided to give it a try, on what I thought would be hard one 'Modeling Operational Risk Capital The Inconvenient Truth" published in the Journal of Operational Risk in 2015
https://lnkd.in/gs6AJbr3
I personally believe it is a very important paper with deep implications for Operational Risk Capital - but I would think that, wouldn't I?
But it is after all an academic paper in a 'pay to read' somewhat obscure journal , so not many people even OP Risk practitioners will ever read it - but such is life!
Anyway I submitted the full text of the paper (warts and all) to Google NotebookLLM and asked for a podcast output (admittedly expecting little).
I was blown away, first (as a technologist) by the sheer smarts of the technology (Wow!) and then the quality of the podcast output which, if I say so myself, nailed the arguments made in the paper. I definitely could NOT have said better myself for a lay audience. It even adds captions!
I have definitely NOT become an AI convert, but have to admire the output. This admittedly early version of the LM program will definitely not create the output (separate research and analysis are needed) but could/will help distribute the research.
Please see further comments below (coming soon) on how researchers might proceed and some of the ethical dilemmas involved
Enjoy (or not)
#operationalrisk
PS had to convert LM output (.wav) to Linkedin (.mp4) so something may have been lost. I assume that will get fixed in time.
Please just press arrow in the black box below