Tag: hallucinations

Study finds ChatGPT-5 is wrong about 1 in 4 times – here’s the reason why

GenAI:N3
A digital, transparent screen is displaying a large, bold '25%' error rate in red, with a smaller '1 in 4' below it. A glowing blue icon of a brain with a gear inside is shown on the left, while a red icon of a corrupted data symbol is on the right. A researcher in a lab coat is looking at the screen with a concerned expression. The scene visually represents the unreliability of a generative AI. Generated by Nano Banana.
While generative AI tools like ChatGPT-5 offer powerful capabilities, new research highlights a critical finding: they can be wrong as often as one in four times. This image captures the tension between AI’s potential and its inherent fallibility, underscoring the vital need for human oversight and fact-checking to ensure accuracy and reliability. Image (and typos) generated by Nano Banana.

Source

Tom’s Guide

Summary

A recent OpenAI study finds that ChatGPT-5 produces incorrect answers in roughly 25 % of cases, due largely to the training and evaluation frameworks that penalise uncertainty. Because benchmarks reward confident statements over “I don’t know,” models are biased to give plausible answers even when unsure. More sophisticated reasoning models tend to hallucinate more because they generate more claims. The authors propose shifting evaluation metrics to reward calibrated uncertainty, rather than penalising honesty, to reduce harmful misinformation in critical domains.

Key Points

  • ChatGPT-5 is wrong about one in four responses (~25%).
  • Training and benchmarking systems currently penalise hesitation, nudging models toward confident guesses even when inaccurate.
  • Reasoning-focused models like o3 and o4-mini hallucinate more often—they make more claims, so they have more opportunities to err.
  • The study recommends redesigning AI benchmarks to reward calibrated uncertainty and the ability to defer instead of guessing.
  • For users: treat AI output as tentative and verify especially in high-stakes domains (medicine, law, finance).

Keywords

, , , ,

URL

https://www.tomsguide.com/ai/study-finds-chatgpt-5-is-wrong-about-1-in-4-times-heres-the-reason-why

Summary generated by ChatGPT 5

GenAI News

Are students really that keen on generative AI?

GenAI:N3
In a collaborative workspace, a male student holds up a tablet displaying generative AI concepts, including a robotic arm, while a question mark hovers above. Another male student gestures enthusiastically, while two female students at laptops show skeptical or thoughtful expressions. A whiteboard covered with notes and diagrams is in the background. The scene depicts students with mixed reactions to generative AI. Generated by Nano Banana.
As generative AI tools become more prevalent, the student response is far from monolithic. This image captures the varied reactions—from eager adoption to thoughtful skepticism—as students grapple with the benefits and implications of integrating these powerful technologies into their academic and creative processes. Are they truly keen, or cautiously optimistic? Image generated by Nano Banana.

Source

Wonkhe

Summary

A YouGov survey of 1,027 students shows strong disapproval of using generative AI for assessed work: 93% say creating work using AI is unacceptable, 82% extend that to using parts of it. While many students have used AI study tools (summarising, finding sources, etc.), nearly half report encountering false or “hallucinated” content from those tools. Most believe their university’s stance on AI is too lenient rather than overly strict, and many expect that academic staff could detect misuse. There are benefits reported—some students think their grades and learning outcomes improved—but overall confidence in AI’s reliability and appropriateness remains low.

Key Points

  • 93% of students believe work created via generative AI for assessment is unacceptable; 82% say even partial use is unacceptable.
  • Around 47% of students who use AI study tools see hallucinations or false information in the AI’s output.
  • 66% believe it likely their university would detect AI-generated work used improperly.
  • Many students report learning and grades that are “slightly more or about the same” when using AI tools.
  • Opinion among students: many are not particularly motivated to use AI for cheating; more often they use it in low-stakes or supportive ways.

Keywords

, , , ,

URL

https://wonkhe.com/wonk-corner/are-students-really-that-keen-on-generative-ai/

Summary generated by ChatGPT 5

GenAI News

How are UK students really using AI?

GenAI:N3
A five-panel mosaic shows diverse UK students interacting with AI. Top row: a Black female student drinks coffee while using a laptop with glowing AI interfaces in a library; a male student studies a book with AI visuals; a male student in a lab coat uses a holographic AI screen with a UK university building in the background. Bottom row: a female student writes while looking at AI data; another male student explains a holographic AI model to a small group of students. The overall image depicts widespread AI integration in student life. Generated by Nano Banana.
From research assistance to code generation and personalised study aids, AI is subtly and overtly reshaping how UK students approach their academic work. This mosaic illustrates the multifaceted ways technology is being integrated into learning, posing questions about its true impact and future trajectory in higher education. Image generated by Nano Banana.

Source

Times Higher Education

Summary

About two-thirds (66 %) of UK undergraduate students report using AI for work or study, with one in three doing so at least weekly. ChatGPT is by far the most popular tool. Universities vary in how they regulate or guide AI use: some discourage it, others provide boundaries but little training, few actively teach ethical AI practice. Most students use AI for understanding concepts, summarising content, or identifying sources, while a smaller but significant share admits to using AI to create or partially create graded work. Many believe AI has helped their grades, though others see little or negative impact. Clearer guidance and teaching around ethical, effective AI use are needed.

Key Points

  • 66% of UK undergrads use AI for degree-related work; 33% use it weekly.
  • The most common applications: explaining difficult concepts (81%), summarising sources (69%), finding sources, and improving existing work.
  • About 14% of AI-using students confess to using AI in graded work in ways that could be cheating (creating or heavily editing), which is ~9% of all students.
  • 47% of AI-using students report frequently encountering “hallucinations” (incorrect or false information) from AI tools.
  • Universities’ policies are mixed: some actively discourage use; many simply warn; only a minority proactively teach students how to use AI well and ethically.

Keywords

, , , , ,

URL

https://yougov.co.uk/society/articles/52855-how-are-uk-students-really-using-ai

Summary generated by ChatGPT 5

GenAI News