GPT-5 Hate! Is it really that bad? Let's take a closer look!

Quick Overview

GPT-5's rollout has been met with significant user criticism regarding short, robotic replies, reduced personality, and a lack of nuanced reasoning, often requiring "prompt engineering." Despite these issues, a recent poll indicates a mixed reception, with roughly equal numbers finding it "good" or "meh," and a small percentage deeming it "bad." OpenAI acknowledges user feedback and is working on improvements, but the rollout has been bumpy, with some users preferring older models.

Key Points: GPT-5 is criticized for short, robotic replies and a less personalized, "colder" tone compared to previous models. Users report that GPT-5 often requires "prompt engineering" to achieve nuanced reasoning and avoid generic or shallow answers. A significant number of users find GPT-5's responses to be factually incorrect or outdated due to training data limitations. Some users believe GPT-5's perceived cutbacks in features like output length and context window are aimed at cost-saving for OpenAI rather than user benefit. There is a notable decline in GPT-5's "agentic" abilities, impacting its planning and multi-step reasoning capabilities. A user poll indicates a mixed reception, with roughly equal numbers finding GPT-5 "good" or "meh," and a small percentage calling it "bad." OpenAI acknowledges user feedback and is working on improvements, including offering more customization options.

Context: The video discusses the widespread user dissatisfaction and critical feedback regarding the performance and features of GPT-5, an AI model developed by OpenAI. It references online discussions, particularly on Reddit and Twitter, and a user poll to illustrate the general sentiment towards the new model.

Detailed Analysis

The video discusses the mixed reception and user complaints surrounding GPT-5, highlighting several key areas of criticism. Firstly, many users report that GPT-5 provides shorter, less nuanced, and more "robotic" replies compared to previous models, often adopting a colder, "AI-stylized" tone that reduces engagement and warmth. This is compounded by a perceived decline in "agentic" abilities, meaning GPT-5 struggles with planning and multi-step reasoning, often requiring users to "engineer" prompts for better output. Users also note that GPT-5 fails to provide accurate information on current events or basic facts, sometimes due to outdated training data. Furthermore, some feel GPT-5 is more about cost-saving for OpenAI than genuine advancement, with features like performance throttling and shorter outputs. Despite these criticisms, a poll shows that of the users who have tried GPT-5, about 32% find it "good," 31% find it "meh," and only 7.4% find it "bad," with 24.6% having "no opinion/haven't used it." The presenter acknowledges the validity of the complaints but also points out that some of these issues, like the need for prompt engineering, have been present in previous models. The video also touches upon Sam Altman's statements acknowledging that OpenAI underestimated how much users cared about GPT-4o's personality and capabilities, and that they are working on offering more customization. However, the overall sentiment conveyed is that GPT-5's rollout has been disappointing for many, especially power users who relied on previous models' strengths.