Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Researchers call ChatGPT Search answers ‘confidently wrong’

Dec 04, 2024 - digitaltrends.com
A recent study by Columbia’s Tow Center for Digital Journalism has found that OpenAI's ChatGPT Search struggles to provide accurate answers to users' queries. The study, which evaluated the accuracy of the AI in identifying the original source of specific quotes, found that ChatGPT returned partially or entirely incorrect responses on 153 occasions. In some instances, the AI misattributed articles or linked to third-party websites that had copied the news article. OpenAI has disputed the study's results, attributing them to flawed testing methods.

In other news, OpenAI is set to launch a new AI agent, code-named Operator, in January, which will enable ChatGPT to take action on the user's behalf. However, reports suggest that the next-generation Orion model of ChatGPT may not live up to the hype, with anonymous OpenAI employees claiming it shows only a "far smaller" improvement over its predecessor, GPT-4. Despite these concerns, ChatGPT's usage continues to grow, with its monthly visits now rivaling those of Google Chrome.

Key takeaways:

  • A study by Columbia’s Tow Center for Digital Journalism found that ChatGPT Search struggles with providing accurate answers to users’ queries, often giving incorrect responses or misattributing sources.
  • OpenAI is set to launch a new AI agent in January, code-named Operator, that will enable ChatGPT to take action on the user's behalf.
  • OpenAI's next-generation Orion model of ChatGPT may not be as advanced as expected, with anonymous employees suggesting it shows only a small improvement over its GPT-4 predecessor.
  • ChatGPT is experiencing significant growth, with 3.7 billion worldwide visits, rivaling browser market share and comparable to Google Chrome's estimated 3.45 billion monthly users.
View Full Article

Comments (0)

Be the first to comment!