OpenAI's GPT-5 rollout is not going smoothly

OpenAI’s GPT-5 rollout is not going smoothly

OpenAI’s recent launch of GPT-5, its eagerly awaited AI model, has encountered several challenges following its debut. Shortly after the announcement, OpenAI CEO Sam Altman acknowledged that the rollout was “more bumpy than we hoped for.” Initial feedback from users indicates that GPT-5 has struggled with basic problem-solving tasks that previous models, including GPT-4o, typically handled more adeptly. Reports include instances where GPT-5 provided incorrect answers to straightforward math problems.

In addition to its mathematical inaccuracies, users noted that GPT-5 did not perform well in evaluating presentation charts from its own launch event. While GPT-5 has been reportedly shown to outperform competitors in coding tasks according to OpenAI’s internal benchmarks, real-world applications suggest that other models, such as Anthropic’s Claude Opus 4.1, may deliver better performance for particular tasks.

Concerns about OpenAI’s internal safety measures have also arisen, with a report indicating gaps in areas such as business alignment and vulnerability to certain attacks. Public sentiment appears mixed; a poll conducted on social media platform X revealed a significant number of respondents feeling unimpressed with GPT-5, describing it as “kinda mid.” Comparisons to previous models, including the deprecation of older versions, have led to user dissatisfaction.

The introduction of a new feature, an automatic routing system for model modes, seems to be contributing to confusion, with reports of users defaulting to less effective settings. This raises questions about the efficacy and reliability of the new model.

As OpenAI faces mounting competition from firms like Google and emerging AI technologies, the reception of GPT-5 underscores challenges in meeting user expectations while maintaining overall performance. Although early user experiences may evolve, the current outlook does not appear as favorable as with previous model launches.

Source: https://venturebeat.com/ai/openais-gpt-5-rollout-is-not-going-smoothly/

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top