Monday, April 22, 2024

Half of consumers experience AI bias

Share

Although 75 percent of people think chatbots are getting better at managing toxic or inaccurate responses, 50 percent have still experienced bias and 38 percent have seen inaccuracies, according to a new report.

The survey of over 6,300 global consumers, software developers and digital quality testing professionals, carried out for testing and digital quality company Applause, finds 91 percent of respondents have used chatbots to conduct research, and 33 percent of those use them for research daily.

Since it was first to market it’s not surprising that ChatGPT has been used the most (91 percent), followed by Gemini (63 percent) and Microsoft Copilot (55 percent). Other chatbots have been used by under a third of users: Grok (32 percent), Pi (29 percent), Perplexity (24 percent), Claude (23 percent) and Poe (21 percent).

Interestingly 38 percent of respondents indicate that they use different chatbots depending on the specific task. 27 percent of respondents also say they have replaced one chatbot with another due to performance.

Of the 1,539 respondents using Gen AI for software development and testing, the most common applications are writing or debugging code (51 percent), test reporting (48 percent), building test cases (46 percent) and building apps (42 percent). GitHub Copilot is the most popular tool for coding assistance (41 percent of respondents), followed by OpenAI Codex (24 percent of respondents).

There are still concerns about privacy though, 89 percent of respondents are concerned about providing private information to chatbots, and 11 percent say they would never provide private information.

“It’s clear from the survey that consumers are keen to use Gen AI chatbots, and some have even integrated it into their daily lives for tasks like research and search. Chatbots are getting better at dealing with toxicity, bias and inaccuracy — however, concerns still remain. Not surprisingly, switching between chatbots to accomplish different tasks is common, while multimodal capabilities are now table stakes,” says Chris Sheehan, SVP strategic accounts and AI at Applause. “To gain further adoption, chatbots need to continue to train models on quality data in specific domains and thoroughly test across a diverse user base to drive down toxicity and inaccuracy.”

You can read more on the Applause blog.

Image credit: BiancoBlue/depositphotos.com

Read more

Local News