Users haven’t been impressed with OpenAI’s new flagship large language (LLM) model, GPT-5. Despite claims of “PhD-level intelligence”, examples of ChatGPT’s basic errors have flooded the internet, as users demand the return of older models.
Via a news post on its website, OpenAI claimed GPT-5 represented “a significant leap in intelligence over all our previous models”. Other lofty marketing statements ranged from the model being “more useful for real-world queries” to “featuring state-of-the-art performance across coding, math, writing, health, visual perception, and more.”
In reality, ChatGPT users found it couldn’t answer basic questions correctly. Unable to spell “Northern Territory” and incorrectly asserting that “blueberry” has three “b”s, GPT-5 also stumbled on queries about geography.
Since the launch of GPT-5 a few days ago, the popular ChatGPT and OpenAI subreddits have been flooded with all sorts of colourful and snarky comments about the LLM’s supposed lack of upgrade.
Controversially, OpenAI removed the ability for subscribers to choose previous models, leaving GPT-5 as the only option. That decision has since been reversed, granting subscribers to ChatGPT’s Plus tier access to the older GPT-4o model.
“We for sure underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways,” OpenAI CEO Sam Altman wrote on X.
Altman faced criticism during a recent Reddit AMA, including perceptions that the newer model was less friendly than GPT-4o. It’s highlighted growing concerns over some users’ reliance on LLMs for companionship and health advice, while OpenAI touts the new model as its “best model yet for health-related questions”.
Limitations on model choices and criticisms of performance led tech commentator Ed Zitron to label OpenAI’s latest move as an example of the “industry-wide enshittification of generative AI”.
With OpenAI yet to hit profitability, the race is on to win users over following a rocky launch.
The post Is GPT-5 a flop? ChatGPT users spark backlash against new model appeared first on GadgetGuy.
0 (mga) komento:
Mag-post ng isang Komento