kxchrischan

Verified purchaser

Deals bought: 47Member since: Jul 2018
1 stars
1 stars
Jul 1, 2024

Worst Buying Experience

I saw a lot of good reviews, and I don't know if they are real or not, but I am skeptical.

Firstly, this tool is very credit-hungry, but I accepted it as they may need to sustain for long-term business. What really pissed me off is that they can't even handle a simple mathematical equation, just to sum up all the numbers. Even when I tried GPT-4, their answers were inconsistent. I'm not making this up; I took screenshots of the responses and sent them to their support. They gave a standard answer that they would reflect on it, but no one cared to reply to me about the problem.

Next, I uploaded 10 files to extract data using GPT-4 (I tried other models as well). While I understand this might be a bit complex, they never gave me the complete data. Each time, I had to prompt them about why the data was incomplete, and only then did they start looking at it. Bear in mind, each prompt took my credits away.

Founder Team
Pratyush_Merlin

Pratyush_Merlin

Jul 3, 2024

Regret your poor experiance. Typically for mathematical queries Code Interpreter is the right feature to be used. Please don't do that on GPT-4 itself. On incomplete data for 10 files, it mainly happens as when the input is too large the model becomes self-constrained on output too which makes it hard to respond fully in one message. This is because there is a 16k token limit on models and that gets divided between input and output tokens. So if the input is too large, the output becomes self constained on output. As models keep expanding token limits and reducing costs, we will be able to generate larger outputs even with smaller inputs.

Helpful?
Share
Ratings