Skip Ribbon Commands
Skip to main content

​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

 

‭(Hidden)‬ Catalog-Item Reuse

A Real Example of the Power and Limitations of Generative AI

A generative artificial intelligence (AI) chatbot was asked to write a quick overview of the types of AI. Here’s how it went.
Sponsored by
a real example of the power and limitations of generative ai

Generative artificial intelligence (AI), like OpenAI's ChatGPT and its numerous cousins, have entered the vernacular. While these types of AI, also known as large language models (LLMs), can answer questions and generate copy in seconds, they do have limitations.

Claude, a ChatGPT-like program from Anthropic AI, was given a copy of Independent Agent magazine's cover story on AI and asked to generate a quick overview of the types of AI based on the context of the article and its knowledge. Below is that overview, straight from the mouth of Claude itself—almost:

Artificial intelligence (AI) is transforming industries across the board, including insurance. There are a few key AI technologies that are especially relevant for independent insurance agents:

  • Machine learning algorithms analyze data to detect patterns, predict outcomes, and make recommendations. This enables fraud detection, automated policy pricing, and identifying customer churn risk. (Independently verified source: ibm.com/topics/machine-learning)
  • Natural language processing (NLP) focuses on understanding and generating human language. NLP powers chatbots to respond to customer FAQs and summarizes documents like claims reports. (The only link Claude provided that worked: marutitech.com/artificial-intelligence-in-insurance)
  • Computer vision analyzes visual data like scanned documents, forms, and images to extract information. It can compare policy documents to spot coverage differences. (Independently verified source: ibm.com/topics/computer-vision)
  • Robotic process automation (RPA) streamlines repetitive digital tasks like data entry and extraction. RPA pulls info from documents faster than humans. (Independently verified source: uipath.com/rpa/robotic-process-automation)
  • Generative AI creates original content like text, images and videos upon request. This includes large language models (LLMs) like ChatGPT that can generate marketing copy, reports, and other materials. (Independently verified source: research.ibm.com/blog/what-is-generative-AI)
  • Predictive analytics utilizes historical data to forecast future outcomes. It identifies cross-sell opportunities and calculates risk scores. (Independently verified source: sas.com/en_us/insights/analytics/predictive-analytics.html)

The rapid evolution of these AI technologies is making them more powerful, accessible, and embedded in common tools. This creates major opportunities for independent agents to use AI to enhance productivity, service and competitiveness.

Claude's original answer didn't include large language models (LLMs), even though the feature mentions them several times. When asked where LLMs fall in this list, Claude inserted it into the “generative AI" definition.

Additionally, Claude was asked to provide sources for every list item. While it did provide URLs, it came with a disclaimer: “Links provided by Claude may not always be valid or up to date. We appreciate your patience as we work to improve link accuracy." That's bad enough, but only one of the six links worked. Five led to “page not found" errors. By the time of publication, Anthropic may have figured that out.

It was also surprisingly difficult to find the sources Claude claimed the information was from. Ultimately, with almost all, independent sources had to be used to cross-reference Claude's definitions. However, when ChatGPT was asked to generate a similar list, it flat-out said it couldn't provide real-time sources, so props to Claude for trying.

One other note: Claude included robotics process automation (RPA). However, most sources categorize RPA as automation and not AI, considering AI and automation as two different animals. Interestingly, when ChatGPT was asked to generate a list, it also included RPA. 

While it did take Claude only seconds to generate well-written explanations complete with independent agent-specific examples it picked up from reading the cover story, there's no denying there's room for improvement—and the need for thorough verification. As they say, check yourself before you wreck yourself.

AnneMarie McPherson Spears is IA news editor. 

17401
Tuesday, November 28, 2023
Technology