Some Catastrophic Failures with AI

- driverless car crashes

- image recognition technology confusing a turtle for a gun

- AI chatbot "Tay" who morphed into a racist troll after being let loose on Twitter for less than a day.

Some other examples demonstrating the concerns about AI include

- the use of AI in the COMPAS algorithms used by some judges in the US when deciding whether a person accused of a crime should be released on bail or not. This program searches through the criminal justice system data to identify characteristics of people who are more likely to re-offend and assigns them a risk rating. Research has found that African-Americans were more than twice as likely as similar white defendants to be classed as medium or high risk, ie this means that they are more than twice as likely to be denied bail. In fact, the data  has highlighted the African-Americans face more police scrutiny, are more likely to receive heavier sentences and are more likely to be convicted of crimes associated with poverty than other ethnic groups. The use of this algorithm ignores the causes of high conviction rates.

- New South Wales Police use algorithms to decide who is placed on the Suspect Target Management Plan

"...over half of those on the list are Aboriginal or Torres Strait Islanders. Given less than 3% of Australia's population is Indigenous, this over-representation suggested significant problems..."

Edward Santow 2018

On the other hand, humans are not perfect

"...A study by Israeli researchers that looked at how our judges decide who was granted bail found that the last prisoner before lunch had an almost zero chance of being released. That's compared to 65% chance first thing in the morning and again straight after lunch..."

Edward Santow 2018

Despite this, AI is getting smarter and we will soon rely on it to shop, interact with government and firms, handle health issues, driving, etc.

"...AI and data that fuels it are also changing the way we do business and often with significant human rights implications. Our personal data is now a currency and smart businesses are already transforming themselves into data dealers......forecast to add 40% to the world's GDP by 2030..."

Edward Santow 2018

Generally people are not comfortable about losing control of their personal data, ie sharing information with third parties

"...Our challenge is to build the right safeguards to protect our rights, while also ensuring  we can nurture the innovations that enhance them..."

Edward Santow 2018

 

Search For Answers

designed by: bluetinweb

We use cookies to provide you with a better service.
By continuing to use our site, you are agreeing to the use of cookies as set in our policy. I understand