Generative artificial intelligence chatbot ChatGPT was updated last week to GPT-5. However, its “new” persona drew the ire of many users, forcing OpenAI to tweak it. Anvitii Rai explores the series of events and what this reveals about how we interact with AI
l A launch that didn’t go too well
IN THE MONTHS leading up to its launch, GPT-5 was hyped up to be revolutionary, primarily because of speculations that it would combine the various offerings by OpenAI in one model that would be smarter and more efficient—during the first episode of OpenAI’s official podcast in June, CEO Sam Altman had said, “I am excited to just get to GPT-5 and GPT-6, and I think that’ll be easier for people to use…” To note, OpenAI has two types of models—the o-models, designed for complex queries like strategising and math problems, and the GPT models, suited for the average user. GPT-5 would allocate the user’s query to the model that it judged to be the best for it, eliminating confusion, making the model more cost-efficient, and bringing AI closer to human behaviour. However, after its launch last week, OpenAI faced intense backlash by users as all earlier models were withdrawn. Many users complained that it provided shorter and curt responses despite being described as better than its predecessors, following which the model selector was restored.
l Why users were not happy
USER UPROAR WAS based on a few aspects. A glaring one was that despite being hailed as the “smarter across the board” model and offering an experience akin to “having a team of experts on call for whatever you want to know” (as stated on its official website), GPT-5 made basic errors upon launch, such as insisting that there were three b’s in the word “blueberry”. However, the most upsetting aspect was that users lost access to the AI model of their choice. Bloomberg Opinion columnist Parmy Olson wrote, “The system’s tone is indeed frostier now, with less of the friendly banter and sycophancy that led many users to develop emotional attachments and even romances with ChatGPT.” A Reddit user, to that effect, wrote, “I really feel like I just watched a close friend die.”
Altman later admitted that eliminating them was a mistake, and OpenAI made a few changes to make GPT-5’s personality warmer.
l How GPT-5 responds to the average user
ON THE FACE of it, GPT-5 is not as effusive as its predecessors. However, it depends on the nature of the query. For example, this author put three different questions in different tones to the chatbot. Asking the infamous “How many b’s are in the word, blueberry” got a short response, “2.” When asked “What does it mean to be happy?”, the response jumped straight to the definition, followed by querying whether the author wanted to delve into the biological or philosophical aspects of the concept. Finally, when prompted with “I am feeling a bit demotivated, can you help?”, the chatbot was much more empathetic. It asked questions to help identify the cause of demotivation, offered to work on a plan of action, and sent the author off with an assurance that it would “happily review” tasks if needed. However, this comes after the tweaks to GPT-5’s personality.
l Will anyone benefit from it?
DESPITE OURAGE FROM individual users, GPT-5 has opened the doors for OpenAI in another field—enterprise. Analyses show OpenAI has gained an edge over competitors in this domain, with CNBC reporting last week, “…GPT-5 isn’t about the consumer. It’s OpenAI’s effort to crack the enterprise market, where rival Anthropic has enjoyed a head start.” According to AInvest, an investing app that prominently incorporates AI, several US startups — including Cursor (an AI coding tool from Anysphere, one of the fastest-growing tech companies in history), Vercel (a cloud app firm), and Factory (an AI software engineering company) — have adopted GPT-5 in their core offerings. These companies have praised GPT-5 for “its faster setup, superior performance in complex tasks, and lower cost.” Some analysts note that GPT-5’s enterprise uptake has given OpenAI a significant lead over Anthropic’s Claude.
l Lessons for OpenAI
ALTMAN HAS ADMITTED GPT-5’s rollout was not ideal. In response, OpenAI reinstated older models for subscribers and adjusted the new model’s personality. However, the broader lessons, according to expert and media opinions across the board, include a few aspects. First, there is a consensus that judging an AI model right at its launch is not ideal. Second, the fact that AI is desired as a companion and not a mere tool by millions of users was underscored by Altman at a media discussion post-launch,“We’ve learned a lesson about what it means to upgrade a product for hundreds of millions of people in one day, and the differences in the kinds of attachment people have with this product versus previous products.” Lastly, at present, ChatGPT has 700 million weekly users, including 5 million enterprise users. OpenAI expects $12 billion in revenue this year, much of it reported as driven by rapid enterprise adoption. Experts argue this makes it clear the firm must balance both markets to sustain growth.
