Refunded - too unreliable
When comparing the answer to ChatGPT or Claude, I often felt like Frank fell short of these other servics, both in depth and quality of reply. Additionally, I got factually wrong information too many times, for basic information such as the name of a company who produces a specific piece of music gear. Big pieces of hardware, not microcomponents (which would have been understandable). Frank got lost in loops such as "so and so makes this interface". Knowing it's not, I asked "are you sure that's the company?" and received a "Sorry, I was wrong it's so and so that makes this interface" - putting another company name (always in the field, but who does not make this specific interface) and when challenged, Frank apologized and replace random company names for like 6/7 times. When it randomly got the right one, I asked it "are you 100% sure this is correct?" and it apologized and changed the name again with a wrong company. On the other hand, the same query on ChatGPT got it right the first time and when challenged it essentially told me "I'm right - it's clearly identifiable information". Very disappointed and requested a refund.
John_FrankAI
Sep 13, 2024Hey there,
The underlying models are GPT3.5, GPT-4 and GPT-4o, which sometimes can provide unreliable results - this is called hallucination and is a common phenomenon for LLM's.
Sorry you had this experience.
Team Frank