Machine Learning Times
EXCLUSIVE HIGHLIGHTS
2 More Ways To Hybridize Predictive AI And Generative AI
  Originally published in Forbes Predictive AI and generative AI...
How To Overcome Predictive AI’s Everyday Failure
  Originally published in Forbes Executives know the importance of predictive...
Our Last Hope Before The AI Bubble Detonates: Taming LLMs
  Originally published in Forbes To know that we’re in...
The Agentic AI Hype Cycle Is Out Of Control — Yet Widely Normalized
  Originally published in Forbes I recently wrote about how...
SHARE THIS:

2 years ago
ChatGPT’s Performance Is Slipping, New Study Says

 
Originally published in Decrypt, July 19, 2023.

UC Berkeley researchers found that ChatGPT has not improved over time, and in fact, may have gotten worse.

ChatGPT exploded onto the scene late last year, dazzling people with its human-like conversational abilities, and the release of latest version prompted a  crypto rally and calls for a pause in development. But according to a new study, the leading AI bot’s skills may actually be on the decline.

Researchers at Stanford and UC Berkeley systematically analyzed different versions of ChatGPT from March and June 2022. They developed rigorous benchmarks to evaluate the model’s competency in math, coding, and visual reasoning tasks. The results of ChatGPT’s performance over time were not good.

The tests revealed a startling drop-off in performance between versions. On a math challenge of determining prime numbers, ChatGPT solved 488 out of 500 questions correctly in March, an accuracy of 97.6%. However, in June, ChatGPT only managed to get 12 questions right, plunging to 2.4% accuracy.

To continue reading this article, click here.

7 thoughts on “ChatGPT’s Performance Is Slipping, New Study Says

  1. As the backrooms game continue to capture the imagination of online communities, game developers, and players alike, the digital exploration of these mysterious spaces serves as a testament to the enduring power of internet-driven storytelling and shared experiences.

     
  2. The recent study highlighting the decline in ChatGPT’s performance is indeed concerning. A drop from 97.6% to 2.4% accuracy in prime number identification between March and June 2023 is significant. This suggests that continuous monitoring and evaluation of AI models are crucial to maintain their reliability.

    For those seeking consistent and dependable AI interactions, platforms like gptde.de offer access to ChatGPT in German without the need for registration. Such platforms can be valuable resources for users requiring stable AI assistance.