[ad_1]
When a person He dies in a car accident in the US. Data about the incident are usually reported to the Highway Traffic Safety Administration. Federal law requires civilian pilots to report flight fires and other incidents to the National Transportation Safety Commission.
The harsh records are intended to provide authorities and manufacturers with better information on ways to improve safety. They helped inspire a multimedia-equipped repository Artificial intelligence events aimed at improving security in much less regulated areas, e.g. autonomous vehicles and robotics. The AI incident database It was launched in 2020 and now has 100 events, among others # 68, the security robot fell into a fountain, and # 16Google’s photo editing services labeled blacks as “gorillas.” Think of it as the AI Shame Room.
There is a database of AI incidents Collaboration on AI, a non-profit created by large technology companies to investigate the disadvantages of technology. The disgrace was started by Sean McGregor, who works as an engineer to learn how to start a Syntiant voice processor. He says it’s necessary because AI allows machines to directly intervene in people’s lives, but a software engineering culture doesn’t promote security.
“I will talk to my engineers often and they will have a very clear idea, but you have to say‘ have you thought about how you are doing dystopia? “, Says McGregor. He hopes to stick the event database as a carrot and to tech companies, offering a way of public accountability that will keep companies off the list while helping engineering teams create AI implementations that are less likely to go wrong.
The database uses a broad definition of AI incidents as “a situation in which AI systems caused or nearly caused real-world damage.” The first entry in the database alleges that YouTube Kids showed adult content, including sexually explicit language. Last, # 100, Is a problem in which a problem in the French welfare system can wrongly determine that people owe money to the state. These include accidents involving autonomous vehicles, for example Uber’s deadly incident in 2018, and wrongful arrest for failure automatic translation or facial recognition.
Anyone can submit an article to the AI disaster catalog. McGregor supports the increases for the time being and has a huge delay in processing, but hopes to eventually maintain the database itself and be an open source project with its community and curatorial process. One of them his favorite events It is an AI agent triggered by a jaywalking detection system that detects a face in Ningbo (China), accusing the woman of appearing wrong in an advertisement next to a bus.
Of the 100 incidents recorded so far, there are 16 involving Google, more than any other company. Amazon has seven, and Microsoft has two. “We are familiar with the database and fully support the role of the partnership and the goals of publishing the database,” Amazon said in a statement. “Gaining and maintaining the trust of our customers is our highest priority, and we have designed rigorous processes to continually improve our services and customer experiences.” Google and Microsoft did not respond to comments.
The Georgetown Center for Security and Emerging Technologies is trying to make the database stronger. Entries are currently based on media reports such as incidents 79, which he mentions Cable reporting according to the algorithm for calculating renal function, according to the design assessing the severity of the disease in black patients. Georgetown students are working to create a helpful database that includes details of an event, such as whether the damage was intentional or not, and whether the problem algorithm acted autonomously or with human input.
Helen Toner, strategy director at CSET, said the exercise informs research into the potential risks of AI accidents. Also, the database shows that it may be a good idea for lawmakers or regulators who see AI rules to do something about reporting incidents, similar to an airplane.
EU and US officials have shown a growing interest in regulating AI, but the technology is so diverse and widespread that it is important to work on clear rules that will not become obsolete quickly. a tremendous job. Recent EU proposals they were accused of excessive reach, techno-illiteracy and numerous gaps. Toner says reporting on AI accidents can help discuss land policies. “I think it would be wise to send them real-world opinions on what we’re trying to prevent and what kind of things are going wrong,” he says.
[ad_2]
Source link
