After subscription to ChatGPT Go, mathematical and scientific skills become shitty?

Title: Evaluating the Impact of Subscribing to ChatGPT Go on Mathematical and Scientific Performance

Introduction

With the increasing popularity of AI-powered tools like ChatGPT, many users turn to subscription services such as ChatGPT Go to enhance their experience and capabilities. However, some users have reported a decline in the AI’s performance in specific domains, notably in mathematical and scientific tasks. This article explores these concerns, examines potential causes, and offers guidance for users seeking consistent AI performance.

User Experience Concerns

A recent user experience highlights issues with the quality of responses from ChatGPT Go after subscribing, especially when using a new account. The user observed that the AI’s performance in generating structured scientific or mathematical content was subpar compared to expectations. Specifically, when requesting the creation of a series of items (e.g., 1-80 questions), the AI produced incomplete outputs—such as only generating questions 1-20 for Part A, skipping to questions 31-50 for Part B, and then continuing with questions 56-80 for Part C. This inconsistent and incomplete output suggests potential limitations or issues in the AI’s ability to handle complex, structured tasks.

Potential Causes

Several factors could contribute to these observed issues:

1. Account Initialization and Data: Creating a new account may require the AI to undergo a form of recalibration or “learning phase,” during which its performance may temporarily decline or be less tailored to the user’s specific requirements.

2. Model Variability: Updates or differences in the underlying AI model for new accounts or subscription tiers might influence response quality, especially in specialized domains like mathematics and science.

3. Prompt Engineering: The way prompts are structured significantly affects output quality. Vague or complex instructions without clear guidance could lead to incomplete or inconsistent responses.

4. System Limitations: AI models have limitations in maintaining long, multi-part outputs consistently. Challenges in generating large, multi-step tasks could result in skipped items or incomplete lists.

Guidance for Users

If you experience similar issues, consider the following strategies:

• Refine Your Prompts: Be explicit and detailed in your instructions. For example, specify that the AI should generate all questions sequentially from 1 to 80, ensuring completeness.

• Break Tasks into Smaller Segments: Instead of requesting all items at once, try requesting smaller sections (e.g., 20 questions at a time) to improve output accuracy.

• **Check for Model Updates