Skip to main content

Here’s why people think GPT-4 might be getting dumber over time

As impressive as GPT-4 was at launch, some onlookers have observed that has lost some of its accuracy and power. These observations have been posted online for months now, including on the OpenAI forums.

These feelings have been out there for a while, but now we may finally have proof. A study conducted in collaboration with Stanford University and UC Berkeley suggests that GPT-4 has not improved its response proficiency but has in fact gotten worse with further updates to the language model.

GPT-4 is getting worse over time, not better.

Many people have reported noticing a significant degradation in the quality of the model responses, but so far, it was all anecdotal.

But now we know.

At least one study shows how the June version of GPT-4 is objectively worse than…

— Santiago (@svpino) July 19, 2023

The study, called How Is ChatGPT’s Behavior Changing over Time?, tested the capability between GPT-4 and the prior language version GPT-3.5 between March and June. Testing the two model versions with a data set of 500 problems, researchers observed that GPT-4 had a 97.6% accuracy rate in March with 488 correct answers and a 2.4% accuracy rate in June after GPT-4 had gone through some updates. The model produced only 12 correct answers months later.

Another test used by researchers was a chain-of-thought technique, in which they asked GPT-4 Is 17,077 a prime number? A question of reasoning. Not only did GPT-4 incorrectly answer no, it gave no explanation as to how it came to this conclusion, according to researchers.

ChatGPT being asked about a prime number.
Image used with permission by copyright holder

The study comes just six days after an OpenAI executive tried to quell suspicions that GPT-4 was, in fact, getting dumber. The tweet below implies that the degradation in quality of answers is a psychological phenomenon from being a heavy user.

No, we haven't made GPT-4 dumber. Quite the opposite: we make each new version smarter than the previous one.

Current hypothesis: When you use it more heavily, you start noticing issues you didn't see before.

— Peter Welinder (@npew) July 13, 2023

Notably, GPT-4 is currently available for developers or paid members through ChatGPT Plus. To ask the same question to GPT-3.5 through the ChatGPT free research preview as I did, gets you not only the correct answer but also a detailed explanation of the mathematical process.

Additionally, code generation has suffered with developers at LeetCode having seen the performance of GPT-4 on its dataset of 50 easy problems drop from 52% accuracy to 10% accuracy between March and June.

To add fuel to the fire, Twitter commentator, @svpino noted that there are rumors that OpenAI might be using “smaller and specialized GPT-4 models that act similarly to a large model but are less expensive to run.”

This cheaper and faster option might be leading to a drop in the quality of GPT-4 responses at a crucial time when the parent company has many other large organizations depending on its technology for collaboration.

Not everyone thinks the study proves anything, though. Some have made the point that a change in behavior doesn’t equate to a reduction in capability. This is acknowledged in the study itself, stating that “a model that has a capability may or may not display that capability in response to a particular prompt.” In other words, getting the desired result may require different types of prompts from the user.

When GPT-4 was first announced OpenAI detailed its use of Microsoft Azure AI supercomputers to train the language model for six months, claiming that the result was a 40% higher likelihood of generating the “desired information from user prompts.”

ChatGPT, based on the GPT-3.5 LLM, was already known for having its information challenges, such as having limited knowledge of world events after 2021, which could lead it to fill in gaps with incorrect data. However, information regression appears to be a completely new problem never seen before with the service. Users were looking forward to updates to address the accepted issues.

CEO of OpenAI, Sam Altman recently expressed his disappointment in a tweet in the wake of the Federal Trade Commission launching an investigation into whether ChatGPT has violated consumer protection laws.

“We’re transparent about the limitations of our technology, especially when we fall short. And our capped-profits structure means we aren’t incentivized to make unlimited returns,” he tweeted.

Editors' Recommendations

Fionna Agomuoh
Fionna Agomuoh is a technology journalist with over a decade of experience writing about various consumer electronics topics…
GPT-4.5 news: Everything we know so far about the next-generation language model
ChatGPT app running on an iPhone.

OpenAI's GPT-4 language model is considered by most to be the most advanced language model used to power modern artificial intelligences (AI). It's used in the ChatGPT chatbot to great effect, and other AIs in similar ways. But that's not the end of its development. As with GPT-3.5, a GPT-4.5 language model may well launch before we see a true next-generation GPT-5.

Here's everything we know about GPT-4.5 so far.

Read more
Newegg wants you to trust ChatGPT for product reviews
AI-generated review on Newegg's website.

Newegg, the online retailer primarily known for selling PC components, has pushed AI into nearly every part of its platform. The latest area to get the AI treatment? Customer reviews.

On select products, Newegg is now showing an AI summary of customer reviews. It sifts through the pile, including the review itself and any listed pros and cons, and uses that to generate its own list of pros and cons, along with its own summary. Currently, Newegg is testing the feature on three products: the Gigabyte RTX 4080 Gaming OC, MSI Katana laptop, and Ipason gaming desktop.

Read more
AI can now steal your passwords with almost 100% accuracy — here’s how
A digital depiction of a laptop being hacked by a hacker.

Researchers at Cornell University have discovered a new way for AI tools to steal your data -- keystrokes. A new research paper details an AI-driven attack that can steal passwords with up to 95% accuracy by listening to what you type on your keyboard.

The researchers accomplished this by training an AI model on the sound of keystrokes and deploying it on a nearby phone. The integrated microphone listened for keystrokes on a MacBook Pro and was able to reproduce them with 95% accuracy -- the highest accuracy the researchers have seen without the use of a large language model.

Read more