Arun Balakrishnan started his career at sea but went back to business school, became an internet entrepreneur and ended up captaining Berkshire Hathaway’s foray into the Indian insurance market.

Ten years ago he founded Insurance technology firm Xceedance and the growth has been rapid.

But that’s just a bit of background because the main purpose of today’s podcast is to talk about Artificial Intelligence (AI) and in particular the generative AI that has exploded onto the world’s consciousness in the past six months.

When something has been this hyped having someone like Arun on the show is an absolute godsend.

Arun is a great explainer and first helps educate me about what the terms bandied around in AI really mean.

Then we start to get to work on decoding what the best applications are going to be in the insurance world.

This is fascinating stuff and we soon get down to the fundamentals of what machines and humans are really best at.

I won’t spoil anything by saying that the humans in insurance really shouldn’t worry about being made redundant by this new technology – there are some things that AI can’t do well, and even if it could, we probably wouldn’t want it to do them for us.

This is far more about improved accuracy and vastly increased productivity. Arun says we should think of it a bit like being allocated a smart intern, apprentice or an indefatigable underwriting assistant.

Arun is a great teacher and I can highly commend this episode to anyone feeling bewildered or daunted as to how to start to engage with this exciting new technological development

NOTES, ABBREVIATIONS AND FURTHER READING

Arun mentions Word2Vec, which comes from ‘word to vector’, a technique whereby words are transformed into a numerical representation – ie. a vector, and is at the heart of the development of this new form of Artificial Intelligence.

Another concept mentioned and worth reading up about further is Zero-Shot classification, which is all about creating a tool that can do a job that it hasn’t been specifically trained to do. An example might be for the AI to have learned a lot about football and then use this insight to classify an article about basketball, upon which it has never been trained.

LINKS & CONTACT

https://xceedance.com/