Dr GPT or: How I Learned to Stop Worrying and Love the AI Part 3
Part 3: A Miniscule Glimmer of Hope
Bing AI for Compiling Relevant Market News and Data
My first two experiments with Bing AI had been somewhat of a failure. (See earlier posts Part 1 and Part 2.) In my third investigation, I thought it could be interesting and relevant to see if it could look at publicly available accounts and assess which insurance companies were exposed to the failures of Credit Suisse and Silicon Valley Bank earlier this year.
I started my research with low expectations; I was not disappointed. Bing AI told me that it was searching the internet for “Insurance companies impacted by the fall of credit Suisse and svb” three times and then told me it couldn’t find anything.
Switching to something more familiar in the actuarial domain, I decided to ask Bing AI about Part VII transfers. (A Part VII transfer, loosely described, is a court-sanctioned legal transfer of some or all of the policies of a company to another, governed by Part VII of the Financial Services and Markets Act 2000.) I first asked it to find a list of recent Part VII transfers and the consultancy that wrote the independent expert report. It informed me it was searching for “recent part VII transfers independent expert report” and let me know that it couldn’t find anything.
Lowering my expectations further, I asked it to find details about one Part VII transfer that took place. It seemed to misunderstand and instead gave me an outline definition of Part VII transfers. In general, that seemed to be simply a cut and paste exercise straight from Wikipedia. I attempted to be more specific by providing it with more information to aid the search and it finally told me about a Life Part VII transfer that took place in 2015. However, it was not able to tell me any more details about the transfer bar very generic details about the amount of and type of policies. When asked who the independent expert was for this report, Bing AI failed to provide an answer.
What was becoming apparent was that there was little benefit in using Bing AI for this type of research. The time it took to get a small piece of information dwarfed the time it would have taken me to search for it myself.
To conclude, for any kind of collating market knowledge or data that is any deeper than surface level, Bing AI adds little more than frustration and wasted effort to the exercise. On reflection, I could have started by searching for something more generic, however, my faith in its ability to deliver remains tainted. Bing AI has shown itself to be neither useful nor efficient to me so far in my investigations. In addition to this, because I knew it to be untrustworthy, I felt the need to double-check every source it sent me and this itself negated any possible remnants of time saved.
- Jack Kelly, Actuarial Analyst at Marcuson Consulting
