Login

News Post

Including evidence in question confuses ChatGPT, lowers its accuracy: study

Economic TimesEconomic Times · 5m
Including evidence in question confuses ChatGPT, lowers its accuracy: study

According to new research, asking AI-powered chatbots health-related questions that include evidence can confuse them and affect their ability to provide accurate answers. The inclusion of evidence in the question may introduce too much noise and lower the chatbot's accuracy. This poses a potential risk as more people rely on online tools, such as chatbots, for health information. The study compared the accuracy of chatbot responses when prompted with evidence-biased questions versus just a question. The results showed a decrease in accuracy when evidence was included in the prompt. Further research is needed to understand the effectiveness of language models in answering health-related questions.

4
View Source

Comments

More Like this

OpenAI’s ‘year of the enterprise’ includes new tools for increasing AI accuracy

OpenAI’s ‘year of the enterprise’ includes new tools for increasing AI accuracy

Meta’s free Code Llama AI programming tool closes the gap with GPT-4

Meta’s free Code Llama AI programming tool closes the gap with GPT-4

Gemini not always reliable in responding to prompts: Google after chatbot's response on PM Modi

Gemini not always reliable in responding to prompts: Google after chatbot's response on PM Modi

Google responds to AI Overviews backlash; unveils corrective measures

Google responds to AI Overviews backlash; unveils corrective measures

Zephr plans to create a "networked GPS" using cell phones

Zephr plans to create a "networked GPS" using cell phones

Use Ola Maps: Bhavish Aggarwal urges developers to exit Google maps

Use Ola Maps: Bhavish Aggarwal urges developers to exit Google maps

Microsoft LASERs away LLM inaccuracies

Microsoft LASERs away LLM inaccuracies

Bengaluru-based Calligo Tech to receive first silicon with posit computing capability this month

Bengaluru-based Calligo Tech to receive first silicon with posit computing capability this month

From Branches To Digital Platforms: How AI Is Simplifying Financial Services

From Branches To Digital Platforms: How AI Is Simplifying Financial Services

Predictions of AI-powered models strictly trial-specific, have no generalisability: Study

Predictions of AI-powered models strictly trial-specific, have no generalisability: Study

Download the medial app to read full posts, comements and news.