AI Evaluation – Beyond the Beltway
Exploring AI’s Word-Finding Capabilities
Can AI Deliver Reliable Answers to Simple Queries?
Recently, while engaged in a game of Wordle, I became curious if there was a five-letter word in English that began and ended with the letter “i.” Instead of using Google, I decided to pose this question to AI tools to see how they would perform.
Initial Inquiry to Microsoft Copilot
I first asked Microsoft’s Copilot for assistance. Surprisingly, the response was rather underwhelming. The AI suggested words that were either incorrect or not in the English language.
Testing ChatGPT’s Knowledge
Next, I turned to ChatGPT for a different perspective. Although it recognized that “igloo” was incorrect, it made other questionable suggestions, such as “imagi,” which is not a recognized English term. It incorrectly noted “India” as one of the answers, alongside minimal hits like “iliwi,” which also isn’t a valid word.
Appraisal of Siri’s Response
I then tried Siri, which redirected back to ChatGPT, but with a slightly altered response, including some irrelevant information.
Discovering Alternative Tools
Eventually, I shifted to another AI, named Grok, which successfully identified “Iceni,” a word I had encountered in a Doctor Who audio drama—albeit, I had forgotten it was an actual term.
Valid Word Suggestions
After my AI interactions, I decided to consult Google and found a few legitimate five-letter words that start and end with “i,” which include:
- Imari – a type of Japanese porcelain.
- Indri – a species of lemur native to Madagascar.
- Issei – a first-generation Japanese immigrant, often to the United States.
Additionally, some other forms were suggested, which may not appear in the dictionary:
- “Iambi” is the plural of “iambus,” but this is borrowed from Latin.
- “Imshi” means “go away” in Arabic, sometimes used by Australians as slang.
According to Dictionary.com, a list includes “iambi” and “imshi” only.
Reflections on AI Performance
I was quite amazed that neither Copilot nor ChatGPT performed better than a simple Google search. Both AI systems provided wrong answers, while Grok managed to uncover “Iceni.” However, I still found several correct entries that Grok did not mention.
This experience highlighted the importance of having basic research skills, especially when relying on artificial intelligence for information. If one were to trust AI for more complex tasks, such as policy-making, it could lead to misguided decisions.
