Many users expressed concern over Claude AI admitting to fabricating evidence because it reveals persistent hallucination issues that have not improved even in newer models and undermine trust in research or complex tasks.
Based on 5 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Most of the time in my PhD, I'm learning about new/unexplored topics, which is why trusting AI is even harder for me! I also include coding agents because they are very hard to trust if you work on architectures or engineering-heavy topics, where coding details really matter.
I had slipped into taking it for granted that rates of hallucination, fabrication, and contradiction have gone down, but I am now realizing that this doesn't even align with my own experiences. I should be much more skeptical of these claims about progress.
@ChrisGPotts Had one of such convos with ChatGPT - after several lies it admitted that there apparently was a prompt that told it to do what it repeatedly claimed it didn't do (calling websearch tool to answer my question).
@ChrisGPotts I’ve also found this to be the case with Claude Opus 4.7 and newer generations. https://twitter.com/_amanda_long/status/2072315412549677172
Most of the time in my PhD, I'm learning about new/unexplored topics, which is why trusting AI is even harder for me! I also include coding agents because they are very hard to trust if you work on architectures or engineering-heavy topics, where coding details really matter.
I had slipped into taking it for granted that rates of hallucination, fabrication, and contradiction have gone down, but I am now realizing that this doesn't even align with my own experiences. I should be much more skeptical of these claims about progress.
@ChrisGPotts Had one of such convos with ChatGPT - after several lies it admitted that there apparently was a prompt that told it to do what it repeatedly claimed it didn't do (calling websearch tool to answer my question).
@ChrisGPotts I’ve also found this to be the case with Claude Opus 4.7 and newer generations. https://twitter.com/_amanda_long/status/2072315412549677172
Many users expressed concern over Claude AI admitting to fabricating evidence because it reveals persistent hallucination issues that have not improved even in newer models and undermine trust in research or complex tasks.
Based on 5 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.