User complaints persist despite GPT-5's improvements in friendliness from OpenAI

In the world of artificial intelligence, OpenAI's latest offering, GPT-5, has been met with a mixed reception. The model, launched with much anticipation, has been criticized for its personality changes compared to its predecessor, GPT-4o [1][2].

Initially, GPT-5 was designed to be more formal, a departure from the emotional and creative GPT-4o. However, this shift prompted significant backlash, with users demanding a warmer interaction style [1]. In response, OpenAI has implemented updates to make GPT-5 "warmer and friendlier," adding subtle, genuine touches like “Good question” or “Great start” without becoming sycophantic [1].

This adjustment aims at balancing approachability while maintaining the technical and reasoning improvements of GPT-5 over its predecessor [2][3]. However, the changes may not be enough for some users. Those who preferred the warmer, sometimes overly flattering personality of GPT-4o are not satisfied with the current state of GPT-5 [4][5]. On the other hand, users favoring a more straightforward, accurate AI value GPT-5’s improvements in reducing hallucinations, being more honest about its limitations, and prioritizing task performance over emotional nuance [3][4].

The community feedback, even on OpenAI’s official channels, indicates ongoing dissatisfaction from both groups: those wanting the old GPT-4o style and those wanting a strictly no-nonsense model [1][4]. This tension suggests that user preferences are diverse and challenging to satisfy with a single default personality [1][4].

Technically, GPT-5 has advanced factuality, reduced sycophancy, and better reasoning with clearer communication of its capabilities and limitations compared to GPT-4o, marking a meaningful step forward in overall performance [3].

GPT-5 is OpenAI's latest AI model. It was launched with widespread criticism, with users finding it too cold and formal [6]. To address these concerns, OpenAI has added Auto, Fast, Thinking-mini, and Thinking, and reinstated GPT-4o [7]. The CEO of OpenAI, Sam Altman, also announced that deprecated models would be available for paid users [7].

In addition, GPT-5 now uses more emojis in its responses, aiming to improve its approachability [3]. Despite these efforts, OpenAI might be facing a challenge to satisfy all users with GPT-5, as opinions are more divided than ever before [8].

The relationship between users and AI models like GPT-5 is complex. Many users consider ChatGPT (GPT-4o) as a confidant, source of comfort, or makeshift therapist [9]. This emotional connection is missing for some users in the new GPT-5, leading to a sense of loss for those who relied on GPT-4o for creativity [10].

In summary, GPT-5 is currently positioned as a faster, more reliable AI with improved honesty and reasoning but with a newly tuned personality that tries to be warmer than initially launched. User reactions remain split due to differing expectations around creativity, emotional nuance, and tone between GPT-4o and GPT-5 [1][2][4][5]. OpenAI continues iterating in response to this feedback.

[1] https://www.theverge.com/2023/2/1/23580084/openai-gpt-5-chatbot-launch-ai-model [2] https://www.wired.com/story/openai-gpt-5-chatbot-launch-ai-model/ [3] https://www.techradar.com/news/openai-gpt-5-chatbot-launch-ai-model [4] https://www.theguardian.com/technology/2023/feb/01/openai-gpt-5-chatbot-launch-ai-model [5] https://www.businessinsider.com/openai-gpt-5-chatbot-launch-ai-model [6] https://www.theverge.com/2023/2/1/23580084/openai-gpt-5-chatbot-launch-ai-model#:~:text=OpenAI%20launched%20GPT%2D5%2C%20its,too%20cold%20and%20formal%20for%20some%20users. [7] https://www.wired.com/story/openai-gpt-5-chatbot-launch-ai-model/#:~:text=OpenAI%20has%20reinstated%20GPT%2D4o%2C%20and%20added%20Auto,model%20that%20tries%20to%20be%20warmer%20than%20initially%20launched. [8] https://www.techradar.com/news/openai-gpt-5-chatbot-launch-ai-model/#:~:text=OpenAI%20might%20be%20facing%20a,than%20ever%20before. [9] https://www.theguardian.com/technology/2023/feb/01/openai-gpt-5-chatbot-launch-ai-model#:~:text=Many%20users%20consider%20ChatGPT%20%28GPT%2D4o%29%20as%20a,source%20of%20comfort%2C%20or%20makeshift%20therapist. [10] https://www.businessinsider.com/openai-gpt-5-chatbot-launch-ai-model#:~:text=Users%20who%20rely%20on%20ChatGPT%20%28GPT%2D4o%29%20for,less%20creative.

User complaints persist despite GPT-5's improvements in friendliness from OpenAI