ArsTechnica, January 4, 2023
A Black man from Georgia was reportedly arrested and held for almost a week based on an incorrect facial recognition match that linked him to thefts he did not commit. According to a report by NOLA.com, Randal Reid, 28, was pulled over by local police on November 25th as he drove on Interstate 20 in DeKalb County, headed to a late Thanksgiving celebration with his mother. The police informed Reid that there was a warrant out for his arrest from Jefferson Parish, Louisiana for theft. However, Reid stated that he had never even been to Louisiana. He was subsequently booked into the DeKalb County jail as a fugitive but was let go on December 1st.
Reid’s lawyer, Tommy Calogero, reported that the Jefferson Parish Sheriff’s Office detectives “tacitly” admitted their error and rescinded the warrant. Calogero believes that the police could have easily determined that Reid was not the perpetrator by checking his height and weight or by making an effort to speak with him or even by asking to search his house for evidence. It is not currently known what specific facial recognition technology was used in this case.
The case began with a June 2022 theft of over $10,000 worth of Chanel and Louis Vuitton purses in Metairie, Louisiana. Reid was ultimately cleared of the crime when a mole on his face, which was absent from the perpetrator captured on surveillance footage, was noticed as a key difference between the two. Additionally, Calogero estimated a 40-pound weight difference between Reid and the thief. This again raises questions about the accuracy of facial recognition technology and the potential for innocent individuals to be wrongly targeted. The Jefferson Parish Sheriff’s office did not respond to requests for information and denied a formal request for the arrest warrant, citing an ongoing investigation.
Read about this incident at the link.
Vice, January 5, 2023
Do Not Pay, an organization known for automating various processes such as fighting parking tickets and canceling unwanted subscriptions, has now created a tool that uses AI to automate customer support calls. The tool uses an AI-generated version of the user’s voice to negotiate refunds for wire transfer fees with banks such as Wells Fargo, as demonstrated in a video uploaded to Twitter by the organization’s founder, Joshua Browder.
In the video, the AI-powered bot is able to navigate the call with a Wells Fargo customer support representative, providing necessary information and effectively arguing for refunds on wire transfer fees. According to Browder, the tool was built using a combination of Resemble.ai, a site that allows users to create their own AI voices, GPT-J, an open-source casual language model, and Do Not Pay’s own AI models for the script. This tool aims to save customers time and effort by eliminating the need for them to spend long periods of time on the phone with customer support. The AI-powered bot is able to handle the conversation in a natural and efficient manner, allowing customers to focus on other important tasks. The organization plans on making the tool publicly available with a range of generic custom voices, but for an additional cost, users will be able to have their own voice.
However, Wells Fargo, in a statement sent after the publication of the article, denied that the call happened as presented in the video. According to Wells Fargo, “It appears the customer’s video recording of this call was edited to look like a bot was interacting with a customer service representative.” But Browder denied this, and stated that the video was not edited except for removing his account number from the start of the clip. The founder added that they called the bank a dozen times to test the flow. The use of AI in customer service is becoming more prevalent, and this tool serves as an example of how AI-powered bots can help customers save time and effort while still getting the desired outcome.
Read more in the article at the link.
Stay up to date with what is happening! Get a first look at news, noteworthy research and worst attacks on AI delivered right in your inbox.
Rise of the machines: 2023 could be the year of artificial intelligence The Globe and Mail, December 22, 2022 As 2023 approaches, the global economy is struggling and businesses are ...
