Skip to Content
logologo
AI Incident Database
Open TwitterOpen RSS FeedOpen FacebookOpen LinkedInOpen GitHub
Open Menu
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse
Discover
Submit
  • Welcome to the AIID
  • Discover Incidents
  • Spatial View
  • Table View
  • List view
  • Entities
  • Taxonomies
  • Submit Incident Reports
  • Submission Leaderboard
  • Blog
  • AI News Digest
  • Risk Checklists
  • Random Incident
  • Sign Up
Collapse

Report 3749

Associated Incidents

Incident 6425 Report
ChatGPT Glitch Disrupts User Interactions with Nonsensical Outputs

Loading...
ChatGPT Is Acting Weird and No One Knows Why
tech.co · 2024

All those strange queries and poems about dogs might have finally caught up to the world's most popular AI chatbot, as users are reporting some seriously erratic behavior from ChatGPT.

ChatGPT and the dozens of generative AI platforms that followed have been far from perfect since their inception. Even the companies that develop these AI chatbots admit that they often suffer from hallucinations, which cause a wide range of AI errors and mistakes.

This goes far beyond simple misinformation, though, with ChatGPT users noting that the platform has “gone crazy” in more ways than one.

According to multiple users who posted interactions on Reddit, ChatGPT is acting very strangely.

One user reported that the AI chatbot started repeatedly responding with “Happy listening,” over and over again, after being asked about jazz music. Another received an unhinged, multi-paragraph rant when asked the simple question,” what is a computer?” on the platform.

Even stranger, another user received a response written in Spanglish, a combination of English and Spanish, without any such prompting whatsoever.

“I really apologise if my last response came through as un unclear or se siente like it drifted into some nonsensical wording… the cogs en la tecla might get a bit whimsical. Muchas gracias for your understanding, y I'll ensure we're being as crystal-clear como l'eau from now on.” – ChatGPT response

Suffice to say, this is not the purpose of ChatGPT, and the generative AI platform is clearly suffering some kind of breakdown. But why?

One of the crazier things about this is that we don't exactly know why ChatGPT is acting so weird. OpenAI has come out with a statement, that honestly doesn't provide much insight.

“We are investigating reports of unexpected responses from ChatGPT. We're continuing to monitor the situation.” – OpenAI announcement

After that, OpenAI came out and said that the “issue has been identified,” but offered no specific explanation for the strange behavior.

Despite the lack of information from OpenAI about why ChatGPT is acting strangely, some experts have deduced a few theories about the potential reasons behind the behavior, and most of them are related to the fact that these kinds of platforms are taking in a lot of information.

“A model like GPT4 is not one single model. It is a learning system that's actually very rapidly changing over time. One big reason why the systems are evolving over time is that they are continuously getting feedback data from millions of users. I think that's one big source of potential changes in the model.” – James Zou, a professor and AI researcher at Stanford University, to Business Insider.

Simply put, ChatGPT and the many AI platforms that have followed are responding to billions of queries and requests and that level of data can have some unexpected consequences.

Read the Source

Research

  • Defining an “AI Incident”
  • Defining an “AI Incident Response”
  • Database Roadmap
  • Related Work
  • Download Complete Database

Project and Community

  • About
  • Contact and Follow
  • Apps and Summaries
  • Editor’s Guide

Incidents

  • All Incidents in List Form
  • Flagged Incidents
  • Submission Queue
  • Classifications View
  • Taxonomies

2024 - AI Incident Database

  • Terms of use
  • Privacy Policy
  • Open twitterOpen githubOpen rssOpen facebookOpen linkedin
  • e1b50cd