The article from Futurism discusses new research indicating that advanced AI chatbots, such as OpenAI’s GPT and Meta’s LLaMA, are becoming increasingly prone to fabricating information despite improving their ability to generate accurate responses. The study published in Nature reveals that while these models can handle complex questions, they also provide more incorrect answers, leading to concerns about their reliability. Researchers note that this trend of “bullshitting” may mislead users into overestimating the accuracy of AI responses, as participants often misjudge the correctness of the answers given.

Editor’s Note: It is a good thing that several journals have already published policies regulating the use of large language models (LLMs) like Chat GPT in the writing of scientific work. We are glad to see that there is a growing recognition of AI hallucinations. Now the question is: why are these LLMs so prone to inventing information? What is in their programs that forces them to do this?

A recent article published by the Computer Methods and Programs in Biomedicine Update says that AI will revolutionize academic writing. According to the authors, AI is beneficial in six areas: idea generation, content structuring, literature synthesizing, data management, editing, and ethical compliance. But if AI is found to be inventing information, how can we trust any information it gives to us? It is already a problem that humans insist on an obsolete materialist view. AI will worsen this, because now, it will create a science that is based on garbage! [Also read Perplexity: Is it better than Google?, Are we passing on our biases to robots?].

As these technologies become integrated into scholarly work, there is a risk that reliance on AI-generated content may diminish the rigor of human inquiry and analysis, as scholars might prioritize efficiency over deep engagement with their subjects. This shift could lead to a generation of researchers less adept at critical thinking and problem-solving, relying instead on AI for answers rather than cultivating their analytical skills. [Read Is artificial intelligence replacing human intelligence?].

In the long run, this trend could erode the foundational principles of intellectual discourse, fostering a society where the depth of understanding is sacrificed for convenience. As we increasingly delegate cognitive tasks to machines, we must preserve our innate capacity for critical thought and ensure that technology serves as a tool for enhancement rather than a crutch that stifles our intellectual growth.

For those who want to see more evidence of AI Lying, we recommend you read the following articles:

• A.I. isn’t making mistakes, it’s lying
• The Most Sophisticated AIs Are Most Likely to Lie, Worrying Research Finds
• Is AI lying to me? Scientists warn of growing capacity for deception
• ChatGPT caught lying to developers: New AI model tries to save itself from being replaced and shut down
• Scheming reasoning evaluations

Read Original Article