Shulou(Shulou.com)11/24 Report--

CTOnews.com June 29, a few days ago, the cyber security company Surfshark released a report on the "AI incident" (AI accident, also known as AI rollover). The report lists the accidents caused by AI since 2010 and analyzes the details of AI accidents.

In the past few decades, especially after entering the 10s of the 21st century, artificial intelligence began to have considerable development, which has affected all aspects of our lives. The ensuing price is that more and more accidents are caused by AI.

▲ figure source Surfshark, the same data below shows that there were 18 AI accidents in the initial phase of 2010-2013.

From 2014 to 2019, that is, the development stage of artificial intelligence, a total of 207cases of AI accidents occurred.

Finally, when artificial intelligence reached the "application stage" in 2020-2023, 293 AI accidents were recorded, an increase of about 15 times compared with 2010-2013.

Among them, there were less than 10 AI accidents each year from 2010 to 2014. The number of accidents has increased sharply every year since 2018, but declined slightly in 2014 and 2019.

With the advent of the GPT-3 architecture developed by OpenAI in 2020 (ChatGPT is based on GPT-3 research and development), the total number of AI accidents began to rise sharply, reaching 91 in 2022.

According to the report, as of May 2023, the total number of AI accidents this year has exceeded half of that of last year, and the growth is expected to continue.

From February 2022 to February 2023, search traffic for chatbots increased by 261%, according to G2, a software review service. Moreover, the three fastest growing software products in the G2 database are all AI products. Considering that the use of AI has skyrocketed in recent years, the significant increase in the number of AI accidents is also "reasonable".

Among these accidents, AI accidents caused by Facebook, Tesla and OpenAI occupy the top three of all known accidents, accounting for 9.3%, 7.7% and 7.5%, respectively.

CTOnews.com found in the report that:

Most of the AI accidents caused by Tesla focus on his Autopilot-assisted driving and FSD fully autonomous driving, which will either brake accidentally or fail to recognize the objects in front of him.

Facebook's AI accident is unable to find and identify violent content and deeply forged audio-visual documents, which can be taken advantage of by lawbreakers.

OpenAI's AI training is easy to cause data infringement and user privacy problems, which often bring all kinds of trouble to the company.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.