In my life, I can recall three specific moments where a piece of technology has completely gobsmacked me.
It was June 2007. I vividly remember waiting on line for several hours on a hot summer day outside the flagship Fifth Avenue New York City Apple Store until I saw my first iPhone. The gorgeous minimalistic design combined with the fact that I had a phone, a computer, a camera, and a music player all-in-one was unbelievable. At that very moment I knew the iPhone would completely revolutionize the world.
In 2019 I test-drove my first Tesla. Sitting in the driver’s seat, I felt like I was in my own personal spaceship. On the highway, the salesperson suggested I pull the lever to engage the auto-pilot setting. I was terrified, but after thinking about it for a few long seconds that really felt like minutes: I did it. Suddenly, I was overtaken by a jolt of adrenaline. I couldn’t believe it. The car was safely driving down the highway, keeping its distance from other cars, and effectively yet cautiously switching lanes. I thought about the endless possibilities and the seemingly inevitable day in the future where driving may not be a human activity.
A few weeks ago, I tried ChatGPT (a commercial solution based upon a large language model - of which many are now in fairly advanced stages of development) and was completely floored . My buddy who was with me at the time told me to ask it any question I wanted. As a huge basketball fan, I asked ChatGPT to “write a poem about why LeBron James is better than Michael Jordan.” Within seconds it generated the following response:
LeBron James is the king of the court,
With skills that can't be beat, he's got the sort.
His stats speak for themselves, they truly shine,
LeBron's game is ahead of Michael's, it's no
longer a sign.
With more points, assists, and rebounds in sight,
LeBron leads the way, day and night.
His efficiency is off the charts, you see,
LeBron's dominance is simply, easily.
So when it comes to who's the best,
LeBron James beats Jordan, he passes the test.
With talent and numbers that can't be ignored,
LeBron James is the basketball king, forever adored.
I was amazed at how an AI language model could quickly pull in data and generate a thoughtful, brilliantly written comparison in what seemed like human writing. The exercise encouraged me to consider some ways new tools like this can be applied to the Property Casualty Insurance space.
For context, ChatGPT has been developed by OpenAI. It generates text responses that are similar to human writing. It uses transformer architecture, a deep learning model, for processing sequential data such as text. It is pre-trained on massive amounts of text data to generate text responses that are close to human language. This allows the tool to learn the patterns and structures of speech so that it can generate high-quality text without additional fine-tuning for specific tasks. The result is a powerful and versatile tool for NLP tasks such as chatbot development, language translation, and text summarization. It is worth noting that the models like these are intended to improve over time. As they ingest more data and receive feedback, their models improve.
I ardently believe that new services like these can be a formidable asset for insurers. Their proficiency in understanding and generating human-like language will facilitate many use cases that can optimize operations, augment customer experience, and mitigate risks.
Probably the most salient near-term application of the technology is in customer service automation. By harnessing the improved language capabilities, insurance companies can create chatbots that can competently respond to common inquiries, such as providing information about policy coverage and facilitating claims processing. This can not only conserve time and resources for both the insurer and the policyholder but also materially elevate the convenience and efficacy of the customer experience.
If I haven’t made it clear by now, I’ll say it: I’m mesmerized by the potential of just how far technologies like this one have come. The possibilities for a step change in performance exist across the value chain, from automating customer service, assisting underwriting, detecting fraud, generating policies, and streamlining the claims process. (And I’m sure I’ve missed some other high-value potential uses cases that will later become apparent). An influential historical figure once said, “There are decades where nothing happens, and there are weeks where decades happen.” It certainly feels like one of those weeks.